I have a skull stripped structural image. How may I begin the recon-all work flow from the segmentation stage? While many stages requires talairach coordinate, should I get the coordinates by running talairach with the original T1 image ( with skull )? however, that would require nuintensitycor to be run, will freesufer have trouble with the intensity during segmentation, because it's not intensity corrected images produced by freesurfer? I actually have made a few different attempts, but I always get a grey matter outline in the WM volume.
if you did not normalize the volume with mri_normalize, you may have trouble using mri_segment to create an accurate wm volume. the segmentation is initially intensity based, expecting mean white matter intensity value of 110.
Also may I use some other software to give freesurfer a segmentation starting point, meaning feeding some segmented images into the wm directory? if so, how should the files be stored?
you could use fast (from FSL) to segment the image. you would need to convert to .img first (make sure to setenv FSLOUTPUTTYPE ANALYZE) and run fast. here's an example of some commands:
mri_convert brain brain.img fast -t1 -c 3 brain.img seg_3.img #the previous line segments the volume into 3 classes (tweak -c vals)
avwmaths seg_3.img -thr 2 -uthr 2 -mul 55 seg_3_110.img #previous line assumes value 2 is wm, we multiply by 55 to get a value of #110. it may be that the value is some other number, threshold and #multiply by an appropriate number to produce a 110(or close) vals
cp brain.mat seg_3_110.mat #copy the .mat (direction info not propagated with ANALYZE)
mri_convert seg_3_110.img wm recon-all -stage2 -subjid <subjid>
if you don't have a talairach transform, you can copy and modify the talairach2 script to use the brain volume.
Thanks in advance,
You're welcome, brian t. quinn