ENH: Support genuine 4D (x,y,z,time) MINC images

[gdevenyi]

MINC volumes with a time dimension now map to a genuine itk::Image<T,4> instead of folding time into components; vector_dimension still maps to components. Fixes the time↔vector round-trip flip and the 3D-only write limit.

Design rationale

Mapping is now by MINC dimension class rather than component count:

• MItime / MItfrequency (MI_DIMCLASS_TIME) → a 4th ITK dimension; tstep→spacing[3], tstart→origin[3], orthogonal direction.
• MIvector_dimension (MI_DIMCLASS_RECORD) → ITK components (deformation grids, DTI, RGB, tensors — unchanged).
• MIxspace/MIyspace/MIzspace → spatial dims.

Apparent dimension order is [time, z, y, x, vector] (slowest→fastest), matching the Read()/Write() hyperslab arrays. SupportsDimension now allows dim < 5.

Behavior change: a MINC file with a time dimension previously read as VectorImage<T,3> (time as N components) and now reads as Image<T,4>. vector_dimension files are unchanged. itkMINCImageIOTest_4D was migrated from VectorImage<float,3> to Image<float,4>; the fixture dti_sample.mnc is a 32-frame time series.

Test results

• ctest -R MINCImageIOTest — 119/119 pass (incl. migrated 4D test, multi-component vector round-trips).
• ITKIOTransformMINC functional tests (itkIOTransformMINCTest, itkMINCTransformAdapterTest) — pass; deformation-grid path (Image<Vector<double,3>,3> ↔ vector_dimension) unaffected.
• Added synthetic Image<float,4> round-trip asserting spacing[3]/origin[3] map to tstep/tstart.

AI assistance

• Tool: Claude Code (Opus 4.8)
• Role: root-cause analysis and implementation
• Contribution: identified the component-count heuristic as the cause of the time↔vector flip; implemented the dimension-class mapping
• All code reviewed and tested locally before committing

[gdevenyi]

@vfonov could you review? This changes MINC time-dimension handling (now genuine 4D itk::Image<T,4>); vector_dimension mapping is unchanged.

[Copilot]

Pull request overview

This PR updates itk::MINCImageIO to treat MINC files with a true time dimension as genuine 4D ITK images (itk::Image<T,4>) instead of folding time into the component count, while keeping MINC vector_dimension mapped to ITK components. This resolves the time↔vector “flip” on round-trips and lifts the previous 3D-only write limitation to allow 4D (3 spatial + time) volumes.

Changes:

• Read path: map MINC time (MI_DIMCLASS_TIME) to an additional ITK dimension (highest), and keep MINC vector_dimension mapped to ITK components.
• Write path: always write ITK components as MINC vector_dimension, and write the 4th ITK dimension as MINC time (reject >4D).
• Tests: migrate the existing “4D” MINC test to itk::Image<float,4> and add a synthetic 4D round-trip that asserts time origin/spacing preservation; add cleanup for the synthetic output.

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
Modules/IO/MINC/src/itkMINCImageIO.cxx	Implements the dimension-class-based mapping for read/write, enabling genuine 4D time series and fixing time/vector ambiguity.
Modules/IO/MINC/include/itkMINCImageIO.h	Updates SupportsDimension to allow up to 4D images (plus degenerate 0 case).
Modules/IO/MINC/test/itkMINCImageIOTest2.cxx	Expands the dimension-support expectation to include 4D.
Modules/IO/MINC/test/itkMINCImageIOTest_4D.cxx	Converts the test to scalar 4D images and adds a synthetic 4D time-geometry/value round-trip check.
Modules/IO/MINC/test/CMakeLists.txt	Ensures the new synthetic MINC output is removed as a temporary test file.

---

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

[greptile-apps]

Greptile Summary

This PR changes MINC time handling to use a real ITK image dimension. The main changes are:

• Allows MINC IO to report support for 4D images.
• Maps MINC time/frequency dimensions to ITK dimension 3.
• Keeps MINC vector_dimension mapped to image components.
• Updates write-side dimension creation and ordering for 4D data.
• Migrates the 4D test to itk::Image<float, 4> with a synthetic round-trip.

Confidence Score: 5/5

This looks safe to merge.

No blocking issues were found in the changed code. The read and write dimension ordering is consistent for scalar, vector, and 4D time layouts, and the updated test includes the headers it uses and cleans up its synthetic output.

No files need attention.

T-Rex Logs

What T-Rex did

• The explicit high-value test plan for the MINC image IO suite executed and completed successfully with all targeted tests passing; no fallback or unrelated tests were run.
• The run log shows that itkMINCImageIOTest_4D and itkMINCImageIOTest2 started and passed.
• The final test summary reported 100% tests passed out of 119.
• The test log was saved for review and verification.

_{Ran code and verified through T-Rex}

Important Files Changed

Filename	Overview
Modules/IO/MINC/include/itkMINCImageIO.h	Updates MINC dimension support from 3D to 4D.
Modules/IO/MINC/src/itkMINCImageIO.cxx	Maps time to an ITK dimension and keeps vector data as components.
Modules/IO/MINC/test/CMakeLists.txt	Adds cleanup for the synthetic 4D test output.
Modules/IO/MINC/test/itkMINCImageIOTest2.cxx	Updates the supported-dimension expectation.
Modules/IO/MINC/test/itkMINCImageIOTest_4D.cxx	Migrates the 4D test to itk::Image<float, 4> and adds time geometry checks.

_{Reviews (1): Last reviewed commit: "ENH: Support genuine 4D (x,y,z,time) MIN..." | Re-trigger Greptile}

[vfonov]

It was done this way to handle DWI acquisitions. And something else, that i don't remember now.
What is the reason for this change?

[gdevenyi]

It was done this way to handle DWI acquisitions. And something else, that i don't remember now. What is the reason for this change?

I was updating https://github.com/gdevenyi/minc-diff-labels to use SimpleITK so I can be agnostic to the filetypes and it fell over on the MINC file saving, so I went to investigate.

If you expand the design "Design rationale" section in the PR top you'll find this maintains the existing implementation for the example DWI tests included in the repo.