 |
 |
Hi Mark, Thanks for taking the time to look over the data I uploaded yesterday. I very much appreciate it. On 18 May 2006, at 17:03, Mark Jenkinson wrote: > Dear Colm, > > The problem with your registrations is that you have not used BET > to remove the non-brain structures from your anatomical image. > As you have low SNR and large FOV it requires some slight tuning > to get BET working, but I found that: > bet 66133.anat anat_bet -c 147 74 98 -f 0.7 > did a reasonable job - good enough for registration anyway. I did try running bet before alright, but it left most of the skull in place. I didn't play with command line arguments. How did you come up with the coords for the centre? Trial and error or something more principled? > > Then, when I ran the first two flirt commands that you had in > your text file, with the newly betted anatomical, I got good > solutions. > I ran bet on the anatomical image using the command line you used. This produced 6133.anat.noskull.nii.gz. Then i tried to normalise against the skull-less anatomical using the following commands: ### Functional /usr/share/fsl/bin/flirt -in /scratch/66133/66133.anat.noskull.nii.gz -ref /usr/share/fsl/etc/standard/avg152T1_brain.hdr -omat /scratch/66133/66133.run1.reg.ed.normalised1.mat -bins 256 -cost corratio -searchrx -180 180 -searchry -180 180 -searchrz -180 180 -dof 12 /usr/share/fsl/bin/flirt -in /scratch/66133/66133.run1.reg.ed.nii -ref /scratch/66133/66133.anat.noskull.nii.gz -omat /scratch/66133/66133.run1.reg.ed.normalised2.mat -bins 256 -cost corratio -searchrx -180 180 -searchry -180 180 -searchrz -180 180 -dof 12 /usr/share/fsl/bin/convert_xfm -concat /scratch/66133/66133.run1.reg.ed.normalised1.mat -omat /scratch/66133/66133.run1.reg.ed.normalised.mat /scratch/66133/66133.run1.reg.ed.normalised2.mat /usr/share/fsl/bin/flirt -in /scratch/66133/66133.run1.reg.ed.nii -ref /usr/share/fsl/etc/standard/avg152T1_brain.hdr -out /scratch/66133/66133.run1.reg.ed.normalised.nii -applyxfm -init /scratch/66133/66133.run1.reg.ed.normalised.mat -interp trilinear ### Anatomical /usr/share/fsl/bin/flirt -in /scratch/66133/66133.anat.noskull.nii.gz -ref /usr/share/fsl/etc/standard/avg152T1_brain.hdr -out /scratch/66133/66133.anat.noskull.normalised.nii.gz -omat /scratch/66133/66133.anat.noskull.normalised.nii.mat -bins 256 -cost corratio -searchrx -90 90 -searchry -90 90 -searchrz -90 90 -dof 12 -interp trilinear But I can't repeat the success you seem to have had. :-( I've uploaded screenshots (number 424955), in fslview, of the anatomical only and functional overlaid on the anatomical. Where have I gone wrong? > The only reason I suggested trying larger search angles or alignment > before is that this can make things easier, but flirt is designed to > work without initial alignment, although if the initial alignment is > nearly 90 degrees or more, then a -180 to 180 search angle should be > used. > > Also, I noted that you have run some clean up on your functional > images. However, I got slightly better registrations when I used the > raw functional images (I used avwroi to extract a single image from > 66133.run1.nii). So I would advise doing the same. Yep, that's correct. The functional is registered, to correct for motion, and then edge detected and then I do the normalisation on the resultant file. So the way to get the best results is to bet the anatomy, and normalise the raw functional data against that and then to the registration and edge detection? In the case where the normalisation is done on the raw functional data, does flirt correctly deal with any relative displacement between the individual slices? Out of curiosity, why do you get better normalisation with the raw functional images than from the cleaned-up images? > > Anyway, the main thing was to use BET on your anatomical image. > Note that this, and other tips for registration, can be found on the > FSL FAQ. Regards, -- Dr Colm G. Connolly School of Psychology and Institute of Neuroscience The Lloyd Building University of Dublin Trinity College, Dublin 2, Éire Tel: +353-1-608-8569 Fax: +353-1-671-3183