Just to clarify, your mask is currently in the subject’s native BOLD/EPI space, and you want to align it to MNI asym space?
To transform data from native space to MNI space,
fMRIPrep uses a vector field deformation computed by composing (1) the affine transform between the subject’s BOLD space and the subject’s anatomical space and (2) the vector-field transform from the subject’s anatomical space to the standard space (in this case, MNI). Thus, you won’t find an BOLD-to-MNI affine transform matrix among the outputs: this would yield a poor alignment in comparison with
fMRIPrep's nonlinear approach.
If your mask is currently in the subject’s native BOLD space, I would instead recommend moving it to MNI space by composing transforms (1) and (2) above.
You can compose and apply them using
antsApplyTransforms to warp the mask into MNI space as follows:
-i </path/to/mask.nii.gz> \
-o </path/to/mask_space-MNI152NLin2009cAsym.nii.gz> \
-r </path/to/reference_MNI_template.nii.gz> \
-n NearestNeighbor \
-t <identifiers>_from-T1w_to-MNI152NLin2009cAsym_mode-image_xfm.h5 \
Note that the above call would change if your mask isn’t binary-valued.