Recently, I sent several emails to ask the same kind of question about building our own template and atlas.
Thanks a lot for the quick replies, they helps me a lot.

 
Currently, I still have a question about building our own atlas using mris_ca_train, sorry to bother again.

In our lab, we have a labeled surface atlas, we'd like to use this atlas instead of using FS's desikan_killiany.2007-06-20gcs.

It is kind of difficult for us to load every subject in our training set to tksurfer, and manually label each ROI based on our ATLAS.
 
>From what I learned from your replies, I did the follwoing steps to creat our own atlas,
 
 
Step# 1  Select a subject, and then send it to oue ATLAS Toolbox for automatic segmentation and parcellation
Step# 2  Sample the volume segmentations onto the surface
Step# 3  Convert the sampled segmentations to an annotation file
 
I applied those three steps to 59 subjects seperately, got .annot file for each of them, then I called the mris_train, my command line is:
 
mris_ca_train -n 2 -t /home/freesurfer/ATLASColorLUT.txt lh sphere.reg myatalas.annot $subjects_list lh_myatlas.gcs
 
My results are shown in the attached .ppt file, my question is:
 
If we looked at Fig. 1(b) and Fig. 2 (b). There are no labels around the ventricl areas in my training set, while when the built atlas was loaded
to average subject, that area was labled.
 
If there are a way to avoid this problem and improve my atlas? Do I use the wrong command ?
 
Thanks!
Guang

Your E-mail and More On-the-Go. Get Windows Live Hotmail Free. Sign up now.