Fix NLST keypoint warping coordinates

[hmeixx]

Fixes #2021

Description

This PR fixes the keypoint coordinate normalization used when sampling the displacement field in learn2reg_nlst_paired_lung_ct.ipynb.

Previously, keypoint voxel coordinates were normalized using the image dimension:

offset = torch.as_tensor(fixed_image.shape[-3:]).to(fixed_keypoints.device) / 2

This corresponds to 2 * x / dim - 1, which does not map voxel indices [0, dim - 1] to [-1, 1].

The normalization is updated to use (dim - 1) / 2, corresponding to:

2 * x / (dim - 1) - 1

align_corners=True is also explicitly specified in F.grid_sample so that the sampling convention is consistent with this normalization and with MONAI Warp.

A small local grid_sample check confirmed that the updated normalization maps voxel centers correctly.

Checks

• Avoid including large-size files in the PR.
• Clean up long text outputs from code cells in the notebook.
• For security purposes, please check the contents and remove any sensitive info such as user names and private key.
• Ensure (1) hyperlinks and markdown anchors are working (2) use relative paths for tutorial repo files (3) put figure and graphs in the ./figure folder
• Notebook runs automatically ./runner.sh -t <path to .ipynb file>

Local PEP 8/format checks passed with runner.sh. Notebook execution was skipped by the existing runner skip pattern.

Summary by CodeRabbit

• Bug Fixes
  • Corrected keypoint displacement sampling to align with PyTorch’s corner-aligned coordinate system, improving registration accuracy and consistency.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Advanced

Run ID: b5a0e6d3-1ae6-4d65-b1c0-9423bde6582f

📥 Commits

Reviewing files that changed from the base of the PR and between b3c6c99 and 2733afb.

📒 Files selected for processing (1)

• 3d_registration/learn2reg_nlst_paired_lung_ct.ipynb

Included review availability: Your plan provides up to 8 included reviews per hour; 7 remain after this review.

---

Walkthrough

The notebook updates keypoint coordinate normalization to use corner-aligned spatial extents and enables align_corners=True when sampling the predicted displacement field.

Changes

Keypoint warping

Layer / File(s)	Summary
Corner-aligned coordinate sampling 3d_registration/learn2reg_nlst_paired_lung_ct.ipynb	Keypoint normalization now uses (dimension - 1) / 2. Displacement sampling explicitly uses align_corners=True.

Priority: ⬇️ Low

Estimated code review effort: 2 (Simple) | ~10 minutes

Change: Bug fix · Severity of issue fixed: Low

Merge Risk: ⚪ Minimal · up to 60b5e

The notebook change aligns keypoint coordinates with displacement sampling without leaving an identified merge-blocking risk.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly and concisely identifies the main change: fixing NLST keypoint warping coordinates.
Description check	✅ Passed	The description explains the coordinate normalization change, the align_corners=True setting, validation performed, and checklist status. It also states why notebook execution was skipped.
Linked Issues check	✅ Passed	The change addresses issue #2021. The keypoint normalization uses (dim - 1) / 2, which maps voxel indices through the maximum index dim - 1 to [-1, 1]. F.grid_sample now sets `align_corners=Tr…
Out of Scope Changes check	✅ Passed	The pull request changes only 3d_registration/learn2reg_nlst_paired_lung_ct.ipynb. The reported changes update keypoint coordinate normalization and sampling alignment, both directly required by iss…
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 0…

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[ericspod]

HI @hmeixx thanks for this fix, it looks good to me.