Hi Maedeh

the number of subjects depends on the variability of whatever you are training. I know Christophe found that a surprisingly small number of subjects (around 15) was sufficient to capture major folding variability, but of course small and rare features might have been washed out and we wouldn't know.

The easiest way I can think of to test your procedure is to train on a single subject, then use mri_ca_label to apply the atlas to that subject. You won't get exactly the same as your training, but it should be awfully close.

cheers Bruce On Sat, 13 Oct 2018, Maedeh Khalilian wrote:

External Email - Use Caution

Dear Freesurfer experts, I have a question, does the number of input subjects matter in using "mris_ca_train" or not? i mean does it affect the results?(maybe the more subjects, the better result) and if yes, what is an approximate number? and the second question is that if there is a way so i can check my .gcs file? for example for reading annotation files u had an m file named "read_annotation.m".it did help me alot. Generally how can i check my .gcs file and evaluate its quality? just by applying it on a subject by "mris_ca_label"  ? i really appreciate you if you can help me,  Kind regards, Maedeh,