Hi Falk,
On May 28, 2009, at 3:50 AM, freesurfer-request@nmr.mgh.harvard.edu wrote:
Message: 4 Date: Thu, 28 May 2009 11:56:51 +0200 From: Falk L?sebrink falk.luese@gmx.net Subject: [Freesurfer] Troubleshooting To: freesurfer@nmr.mgh.harvard.edu Message-ID: <001501c9df7a$9fe15a40$dfa40ec0$@luese@gmx.net> Content-Type: text/plain; charset="iso-8859-1"
Hello,
Im having trouble to generate a correct aseg.mgz for one volume of mine. The nu_noneck.mgz is also incorrect and therefore I think the -careg stage of autorecon2 might fail. Is there a way to do some kind quality check, like the tal-check during autorecon1, to see whether the talairach.m3z is incorrect? Are there some kind of instructions like fixing the bad output from the talairach registration? Or are there some parameters which can be adjusted to get better results?
If none of this is possible in some way, would it be advisable to use the -noaseg flag? Im actually in no need for the subcortical segmentation, it would just be a plus, but the aseg.mgz is needed during the - segmentation stage to remove the cerebellum and brain stem. Maybe there even is another way to remove these structures automatically?
Well, I can offer a very rough method that seems to work with my monkey data (where I always use -noaseg), I never tried it with humans. I edit the wm.mgz and delete two slices worth of midbrain above the upper connection of the cerebellum to the midbrain. It seems important that there are no connections from cerebellum left to white matter (if there are connection above the cut I always delete the cerebellar part of those, but I do not care about the cerebellum currently). Then I pick a point on the pons in the segment above the cut. I use this point, as well as a point picked somewhere in the middle of the corpus callosum to feed into further recon-all steps (see below). This probably is rather unclean, but it seems to work okay, at least with monkey data...
Here my actual script for that step (replace the values in CC_string and PONS_string with your real values picked from tkmedit of wm.mgz): # set the target directory and file_name set targ_dir = ../mri set src_vol = wm.mgz #set brain = brain.mgz set targ_vol = filled.mgz set controlpoints = ../tmp/control.dat set cur_dir = `pwd`
# ATTENTION please specify the cutting planes for the pons and the corpus callosum # define the raw voxel coordinates for CC and PONS set CC_string = "128 110 132" set PONS_string = "128 136 125"#"127 129 112"
# do an initial filling to check the cutting planes set check_filling = 0
cd .. set tmp_subj = 0?????* cd ${cur_dir}
if (${check_filling} == 1) then echo "Testing the CC and Pons coordinates, if the filled volume is okay rerun this script with set check_filling = 0" #mri_fill -CV 127 128 128 -PV 127 154 128 ${src_dir}/${src_vol} $ {targ_dir}/${targ_vol} mri_fill -a ../scripts/ponscc.cut.log -PV ${PONS_string} -CV $ {CC_string} ${src_dir}/wm.mgz ${src_dir}/filled.mgz # check the results, left and right white matter should be colored differently and the cerebellum should be excluded tkmedit -f ${src_dir}/filled.mgz -aux ${src_dir}/T1.mgz exit 0 endif
# 20060224 recon-all -autorecon2-wm \ -subjid ${tmp_subj} \ -cc-crs ${CC_string} \ -pons-crs ${PONS_string} \ -noaseg \ -xopts-use
Thanks in advance,
Regards,
Falk
I hope this helps Sebastian