ENH: Add ngff-zarr multi-resolution registration example by thewtex · Pull Request #375 · InsightSoftwareConsortium/ITKElastix

thewtex · 2026-03-09T12:17:59Z

Add Example 23 demonstrating a workflow that combines ngff-zarr multi-resolution image pyramids with ITKElastix registration:

Convert ITK images to ngff-zarr Multiscales via to_multiscales
Register at a coarse resolution (rigid + affine + bspline)
Convert Elastix results to itk.CompositeTransform
Apply the transform at full resolution in parallel using dask.array.map_blocks with itk.resample_image_filter

Uses itk_image_to_ngff_image and ngff_image_to_itk_image to bridge between ITK and ngff-zarr data representations.

review-notebook-app · 2026-03-09T12:18:05Z

Check out this pull request on

See visual diffs & provide feedback on Jupyter Notebooks.

Powered by ReviewNB

N-Dekker · 2026-03-09T13:40:49Z

@thewtex Please do git rebase main on this PR! It appears 14 commits behind!

By the way, what does "9566-example-demonstr" mean?

Add Example 23 demonstrating a workflow that combines ngff-zarr multi-resolution image pyramids with ITKElastix registration: - Convert ITK images to ngff-zarr Multiscales via to_multiscales - Register at a coarse resolution (rigid + affine + bspline) - Convert Elastix results to itk.CompositeTransform - Apply the transform at full resolution in parallel using dask.array.map_blocks with itk.resample_image_filter Uses itk_image_to_ngff_image and ngff_image_to_itk_image to bridge between ITK and ngff-zarr data representations.

thewtex · 2026-03-24T19:18:27Z

@thewtex Please do git rebase main on this PR! It appears 14 commits behind!

@N-Dekker 👍 done!

By the way, what does "9566-example-demonstr" mean?

This is a unique identifier for the git worktree I am using.

- Reword 'required by Elastix' to 'recommended for Elastix, by default' in markdown and code comment for the float pixel type discussion - Clarify that 'native itk.Image' is in contrast to 'itkwasm.image.Image' when calling ngff_image_to_itk_image with wasm=False - Rename 'resampled_dask' to 'resampled_dask_array' for clarity

N-Dekker

Thanks for looking at my initial comments, Matt! Did you push the changes you made to address them?

N-Dekker · 2026-05-19T13:43:39Z

Maybe beyond the scope of this PR, but "CT_2D_head_fixed.mha" looks a bit warped. Especially when looking at the border of the skull. I see, you introduced the image by 56caa00 (back in 2020).

In MeVisLab ("CT_2D_head_fixed.mha" at the left side, "CT_2D_head_moving.mha" at the right):

Do you still remember where "CT_2D_head_fixed.mha" and "CT_2D_head_moving.mha" came from?

N-Dekker · 2026-05-19T17:32:17Z

+    "def resample_block(block, block_info=None, transform=None, moving_image=None,\n",
+    "                   spacing_itk=None, origin_itk=None):\n",


Why all those =None defaults?

block_info=None, transform=None, moving_image=None, spacing_itk=None, origin_itk=None

It appears that resample_block crashes when doing transform=None, for example.

I would suggest to simply remove all those =None defaults.

N-Dekker · 2026-05-20T12:48:24Z

+    "ngff_fixed = itk_image_to_ngff_image(fixed_image)\n",
+    "ngff_moving = itk_image_to_ngff_image(moving_image)\n",


For readability, can you please rename the following variables, by appending "_image"?

ngff_fixed to ngff_fixed_image

ngff_moving to ngff_moving_image

fixed_coarse to fixed_coarse_image

moving_coarse to moving_coarse_image

fixed_coarse_ngff to fixed_coarse_ngff_image

moving_coarse_ngff to moving_coarse_ngff_image

N-Dekker · 2026-05-20T13:14:41Z

+    "ImageType = itk.Image[itk.F, 2]\n",
+    "if not isinstance(fixed_coarse, ImageType):\n",
+    "    fixed_coarse = itk.cast_image_filter(fixed_coarse, ttype=(type(fixed_coarse), ImageType))\n",
+    "if not isinstance(moving_coarse, ImageType):\n",
+    "    moving_coarse = itk.cast_image_filter(moving_coarse, ttype=(type(moving_coarse), ImageType))"


Why would you do this? (Conditionally casting the images to float) I'm sorry, but I think this part should be removed. First of all, the images returned by ngff_image_to_itk_image are already float images, so itk.cast_image_filter is not called. Secondly, ITKElastix/elastix will already take care of converting the input images to the internal pixel types, automatically. So I think users do not need to "learn" how to do that manually, beforehand.

N-Dekker · 2026-05-20T13:36:24Z

+    "\n",
+    "# Materialize the full-resolution moving image as an itk.Image for resampling.\n",
+    "# Each block function will sample from this image.\n",
+    "moving_itk_full = ngff_image_to_itk_image(moving_full, wasm=False)\n",


moving_itk_full is basically identical to the original moving_image, right? Same size, same pixel values, right? So it's a loss-less round-trip: moving_image ==> itk_image_to_ngff_image ==> to_multiscales ==> images[0] ==> moving_itk_full

Maybe it's worth noting 🤷

N-Dekker · 2026-05-20T14:05:40Z

+    "# Compute the result (triggers parallel execution across chunks)\n",
+    "resampled_array = resampled_dask.compute()\n",


Do I understand correctly that this compute() may internally call resample_block multiple times in parallel, for multiple chunks? And does that actually happen, when running this example?

Is this a relevant use case for dask map_blocks, or is it more like a toy example? The aim is to speed up the resampling by using multi-threading, right? But then, isn't itk.resample_image_filter itself already multi-threaded?

thewtex · 2026-05-20T20:52:09Z

Thanks for looking at my initial comments, Matt! Did you push the changes you made to address them?

@N-Dekker oops, I think I pushed them to the wrong remote -- re-pushed!

thewtex · 2026-05-20T20:52:59Z

Maybe beyond the scope of this PR, but "CT_2D_head_fixed.mha" looks a bit warped. Especially when looking at the border of the skull. I see, you introduced the image by 56caa00 (back in 2020).

This is intentional, an artificially induced warp. I believe these images come from the ITK Software Guide.

N-Dekker · 2026-05-21T11:20:37Z

This is intentional, an artificially induced warp. I believe these images come from the ITK Software Guide.

Ah, thanks, I see similar images now, in InsightSoftwareGuide-Book2-5.4.6.pdf, although not warped:

Still, I would expect the warping to take place on the moving image, during registration (bspline), and no warping of the fixed image 🤷

Update, I see now, the T1 image (before warping) is at https://github.com/InsightSoftwareConsortium/ITKSoftwareGuide/blob/7510e24a4dd8e98b6ee582c2468961f69ca1ec4b/SoftwareGuide/Art/BrainT1Slice.jpg

thewtex requested a review from N-Dekker March 9, 2026 12:18

N-Dekker reviewed Mar 9, 2026

View reviewed changes

Comment thread examples/ITK_Example23_NgffZarrMultiscaleRegistration.ipynb

N-Dekker reviewed Mar 9, 2026

View reviewed changes

Comment thread examples/ITK_Example23_NgffZarrMultiscaleRegistration.ipynb Outdated

thewtex force-pushed the 9566-example-demonstr branch from 09b02cd to 56a87dd Compare March 24, 2026 19:07

thewtex force-pushed the 9566-example-demonstr branch from 56a87dd to 95de05e Compare March 24, 2026 19:19

thewtex marked this pull request as draft March 24, 2026 19:19

N-Dekker reviewed May 18, 2026

View reviewed changes

Comment thread examples/ITK_Example23_NgffZarrMultiscaleRegistration.ipynb Outdated

N-Dekker reviewed May 18, 2026

View reviewed changes

Comment thread examples/ITK_Example23_NgffZarrMultiscaleRegistration.ipynb Outdated

N-Dekker reviewed May 18, 2026

View reviewed changes

Comment thread examples/ITK_Example23_NgffZarrMultiscaleRegistration.ipynb Outdated

thewtex marked this pull request as ready for review May 18, 2026 18:59

N-Dekker requested changes May 19, 2026

View reviewed changes

N-Dekker reviewed May 19, 2026

View reviewed changes

N-Dekker reviewed May 20, 2026

View reviewed changes

thewtex force-pushed the 9566-example-demonstr branch from 95de05e to 96c4f88 Compare May 20, 2026 20:51

		"def resample_block(block, block_info=None, transform=None, moving_image=None,\n",
		" spacing_itk=None, origin_itk=None):\n",

		"ngff_fixed = itk_image_to_ngff_image(fixed_image)\n",
		"ngff_moving = itk_image_to_ngff_image(moving_image)\n",

		"# Compute the result (triggers parallel execution across chunks)\n",
		"resampled_array = resampled_dask.compute()\n",

Conversation

thewtex commented Mar 9, 2026

Uh oh!

review-notebook-app Bot commented Mar 9, 2026

Uh oh!

N-Dekker commented Mar 9, 2026

Uh oh!

Uh oh!

Uh oh!

thewtex commented Mar 24, 2026

Uh oh!

Uh oh!

Uh oh!

Uh oh!

N-Dekker left a comment

Choose a reason for hiding this comment

Uh oh!

N-Dekker commented May 19, 2026

Uh oh!

N-Dekker May 19, 2026

Choose a reason for hiding this comment

Uh oh!

N-Dekker May 20, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

N-Dekker May 20, 2026

Choose a reason for hiding this comment

Uh oh!

N-Dekker May 20, 2026

Choose a reason for hiding this comment

Uh oh!

N-Dekker May 20, 2026

Choose a reason for hiding this comment

Uh oh!

N-Dekker May 21, 2026

Choose a reason for hiding this comment

Uh oh!

thewtex commented May 20, 2026

Uh oh!

thewtex commented May 20, 2026

Uh oh!

N-Dekker commented May 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

N-Dekker May 20, 2026 •

edited

Loading

N-Dekker commented May 21, 2026 •

edited

Loading