Thanks Rick~
Yes, that method sounds reasonable, I've got the mask in tlrc space.
But I met another problem, when I tried to do the non-linear 3D warp using auto_warp.py, for warping the T1 image to TT_27 template.
Here is the code I used :
auto_warp.py -base TT_N27+tlrc -input ${subject}_TPL.nii \
-skull_strip_input no \
-affine_input_xmat AUTO -qw_opts -emask \
SphereMaskInvers+tlrc \
-minpatch 7
Btw, the emask is just a mask of the whole subcortical region.

Here is the error I got:
Warping used up maximum iterations
even after doubling number of iterations to 274.
Convergence might not have been reached.

If you find the alignment poor, try
rerunning the script with a transform
type of fewer parameters (-xform) or by using
an even higher number of iterations (-maxite).
You can also choose to continue with the
current result (dset: __ats_tmp__reg_warpdriveout+orig
by using the option: -OK_maxite

Script will stop ...

#**ERROR Failed

Many thanks,
Sarah