Improve talairach transform and etiv estimation through atlas prealignment

[kdiers]

This PR intends to improve talairach transform and etiv estimation through atlas prealignment.

• Previous talairach transforms have been reported to suffer from registration failures, leading to biased etiv estimates.
• To make talairach registration more robust, we introduce an additional affine prealignment to the mni_icbm152_t1_tal_nlin_asym_09c template, based on neuroreg's centroid segmentation (segreg).
• The talairach_avi routine will then be based on the prealigned image, which gives a better starting point for the registration.
• The final talairach transform will be the concatenation of the transforms returned by the segreg and talairach_avi registrations.

[dkuegler]

Should we delete the prealigned image file after talairach-registration was done? We can keep the ltas, but the image file can be deleted, no?

[kdiers]

Already marking as "ready for review", but still intending to add update to new neuroreg release.

[kdiers]

No further commits planned currently and ready for review now @m-reuter , @dkuegler

[Copilot]

Pull request overview

This PR strengthens Talairach registration robustness (and downstream eTIV estimation) by introducing an affine prealignment step based on neuroreg centroid-based registration, then concatenating that prealignment with the existing talairach_avi transform.

Changes:

• Add neuroreg-based centroid prealignment prior to talairach_avi, and concatenate transforms into the final Talairach transform.
• Plumb through new CLI args (--py, --asegdkt_segfile) to talairach-reg.sh from the main entrypoints.
• Add neuroreg as a dependency and add a centroid JSON resource for the target template.

Reviewed changes

Copilot reviewed 8 out of 9 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
run_fastsurfer.sh	Passes Python command + segmentation path into Talairach registration step.
recon_surf/recon-surf.sh	Passes Python command + segmentation path into Talairach registration step.
recon_surf/talairach-reg.sh	Implements segreg-based prealignment, runs talairach_avi on prealigned image, concatenates transforms.
recon_surf/functions.sh	Adds helper to locate Python-provided console scripts (segreg, lta).
recon_surf/segreg_mni_icbm152_t1_tal_nlin_asym_09c_centroids.json	Adds centroid definition data intended for the prealignment.
requirements.txt	Adds neuroreg to the frozen (container-derived) dependency list.
requirements.cpu.txt	Adds neuroreg to the frozen CPU dependency list.
requirements.mac.txt	Adds neuroreg to macOS dependency list.
pyproject.toml	Declares neuroreg as a project dependency.

Comments suppressed due to low confidence (1)

recon_surf/talairach-reg.sh:85

• talairach-reg.sh now relies on --py and --asegdkt_segfile, but they are not included in the required-argument validation loop (only --dir/--conformed_name/--norm_name are checked). If a caller omits either flag, the script will fail later with less actionable errors. Add explicit validation for python and asegdkt_segfile (or provide safe defaults) before running segreg.

  --py) python="$1" ; shift ;;
  --asegdkt_segfile) if checkfile "$1" ; then exit 1; fi ; asegdkt_segfile="$1" ; shift ;;
  --edits) edits="true" ;;
  --3t) atlas3T="true" ;;
  *) echo "ERROR: Unrecognized argument $key!" ; usage ; exit 1 ;;
esac
done

for arg_spec in "dir=$mdir" "conformed_name=$conformed_name" "norm_name=$norm_name"
do
  arg=$(echo "$arg_spec" | cut -d"=" -f1)
  value=$(echo "$arg_spec" | cut -d"=" -f2)
  if [[ -z "$value" ]] ; then
    echo "ERROR: --$arg is a required argument, but no value was passed!"
    exit 1
  fi
done

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Improves robustness of Talairach registration (and downstream eTIV estimation) by adding an affine atlas prealignment step using neuroreg centroid-based registration, then concatenating transforms for the final Talairach transform.

Changes:

• Add neuroreg as a dependency (pip requirements + pyproject.toml).
• Pass Python interpreter and asegdkt segmentation into the Talairach registration script.
• Update Talairach registration to: (1) compute a segreg-based prealignment, (2) run talairach_avi from the prealigned image, (3) concatenate transforms.

Reviewed changes

Copilot reviewed 6 out of 8 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
run_fastsurfer.sh	Passes --py and --asegdkt_segfile into Talairach registration invocation.
recon_surf/recon-surf.sh	Passes --py and --asegdkt_segfile into talairach-reg.sh from recon-surf pipeline.
recon_surf/talairach-reg.sh	Implements the new neuroreg prealignment + transform concatenation flow.
requirements.txt	Adds neuroreg==0.6.0 to the container freeze requirements.
requirements.cpu.txt	Adds neuroreg==0.6.0 to the CPU container freeze requirements.
pyproject.toml	Adds neuroreg>=0.6.0 to project dependencies.
CerebNet/data_loader/dataset.py	Refactors TypedDict typing for ROI metadata and adjusts construction.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.