Transforming atlas into subject space

Hi neurostars,

I want to apply the Power 2011 atlas (seeds of 264 ROIs in MNI152Lin space) on individual subjects in T1w space.

For this, I first ran fMRIprep with
--output-spaces anat MNI152Lin

Then, I used antsApplyTransformsToPoints to transform my seeds into T1w space using the .h5 file provided by fMRIprep:
antsApplyTransformsToPoints -d 3 -i atlas.csv -o transformed_atlas.csv -t sub-X_from-MNI152Lin_to-T1w_mode-image_xfm.h5

With this approach, many seeds end up completely outside of my BOLD data.
Would this be expected, or are there any fundamental issues with my workflow?

Thank you for your help!

I just found a very similar thread (which I somehow missed previously). I will revive the discussion there: Extracting individual transforms from composite .h5 files (fmriprep) - #4 by jclauneuro