Hi,
I am a little confused about what you're trying to do. If you're using FreeSurfer parcellations, the ROIs you extract will be in subject space though not necessarily in alignment with your data.
You also mentioned MNI 152 ROIs. If you want those resampled to original space then you can do so after you estimate the transform of each subject's data to MNI_152 space. You can then use 3dfractionize to resample the MNI_152 space ROIs to original space.
If it is easier for you, you might as well take all your time series to MNI_152 space and then sample the signals using the existing ROIs. I don't see much of an advantage to creating subject-specific versions of MNI_152 derived ROIs...
cheers,
Ziad