 |
 |
Hi Neils, I applied the registration from the FA images to the b0 brain masks using the following command and still the transformed b0 brain mask has a misplaced center of gravity compared to its corresponding original bo brain mask. applywarp --ref=FA_subject1.nii.gz --in=b0_brain_mask_subject1.nii.gz --warp=FA_subject1_nonlinear_xfm --out=b0_brain_mask_subject1_common --interp=nn any idea? how can I send you the masks? Thank you, Mahmoud On Thu, Jan 12, 2017 at 1:14 PM, Niels Bergsland <[log in to unmask]> wrote: > Hi again, > I think the best thing would be to try to get the brain extraction working > better on your b0 images. If you aren't correcting for geometric > distortions with a field map, then that may potentially have an impact but > it's likely that there is still a way to get a reasonable brain extraction > regardless. If you want to send me a sample b0 image or two, I can try to > take a look. I might not get to it until tomorrow though. > > If you want to proceed with your original idea, then you should just be > able to apply the registrations from the FA images to your b0 masks. > > > On Thu, Jan 12, 2017 at 18:19 Mahmoud <[log in to unmask]> wrote: > >> Hi Neil, >> >> I apologize for the confusion. "bunch of BET images" was not correct. I >> used BET to create a ' binary mask' of the brain tissue from b0 images. BET >> usually does not very well on my b0 images and most of the time parts of >> the brain tissue is chopped off or other not brain stuff is included. >> I have already transformed the FA maps from individual space to a common >> space. Is it a good idea to use those transformation matrices to register >> each binary mask from an individual space to that common space and the >> average them? >> >> Thank you! >> Mahmoud >> >> On Thu, Jan 12, 2017 at 11:53 AM, Niels Bergsland <[log in to unmask]> >> wrote: >> >> Hi Mahmoud, >> You are absolutely right about trying to register a binary mask - and you >> can't directly register it. You need to do the registration using the >> corresponding structural image and then apply the transformations to your >> masks. >> >> I'm still a little bit confused though. You mentioned in the earlier >> email "bunch of BET images" but now it seems like you're referring to >> binary masks? I thought you were referring to brain extracted images. If >> you want to register the binary masks, you should use the corresponding >> structural image. Once you've gotten all of your masks into the same space, >> you can certainly make an average image. >> >> If there's something that I've misunderstood, let me know. >> >> On Thu, Jan 12, 2017 at 5:43 PM, Mahmoud <[log in to unmask]> wrote: >> >> Hi Niels, >> >> Thank you so much for your explanation. >> My concern is that I don't know how much different is registering a >> binary mask from registering a structural image. In the latter, there are >> some many details that might be used for the registration while there is >> not much info in a binary mask so I thought there might be a specific way >> to register binary masks. About the goal of doing this registration: I have >> BET images in individual space. Those images are not perfect. I want to >> register all the images into a common space and then average the registered >> BET images to see what areas of the brain mostly chopped off. >> >> Thank you, >> Mahmoud >> >> On Thu, Jan 12, 2017 at 11:30 AM, Niels Bergsland <[log in to unmask]> >> wrote: >> >> Hi, >> Sorry about that - I read your email too quickly and didn't notice that >> you were not just doing a straight subsampling. There is a section on this >> in the FLIRT FAQ (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FLIRT/FAQ), >> particularly the last paragraph of Can I register to an image but use >> higher/lower resolution (voxel size)?" section. >> >> If you have access to freesurfer, you may find that this is easier to do >> with mri_convert, in that case you can do it like this: >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> mri_convert -vs 2 2 2 -oni 92 -onj 92 -onk 56 INPUT.nii.gz OUTPUT.nii.gz >> >> >> Regarding your other question, it's difficult to say without knowing more >> about exactly what you're trying to do but in general, you would do >> something like an affine registration with FLIRT and then a nonlinear >> registration with FNIRT. There are some basic examples of this on the FNIRT >> webpage. >> >> Hope this helps, >> Niels >> >> >> >> >> >> >> >> >> >> >> >> >> -- >> Niels Bergsland >> Integration Director >> Buffalo Neuroimaging Analysis Center >> 100 High St. Buffalo NY 14203 >> [log in to unmask] >> >> >> >> >> >> >> >>