@mcsnyder Not as written. However, several of the functional subworkflows (BOLD reference estimation, slice-timing correction, head motion correction, fieldmap-based susceptibility distortion correction) do not use anatomical inputs. With a little work, these could be separated from coregistration and resampling. And portions of the confounds workflow do not require anatomical images (or BOLD series resampled in alignment with anatomical images).
Unfortunately, I don’t think we’re going to be able to devote much time to this in the near future, but we would be willing to consider a pull request that accomplishes this separation.