[SID-6921]

Description

Follow-up to #938, whose description explicitly reserves "independent channels_first propagation defects in resample_image, smooth_image, and multicomponent indexing" for later work.

ants.resample_image's multi-component branch reassigns its own image parameter inside the per-channel loop (for image in images:). By the time channels_first=image.channels_first runs after the loop, image refers to the last split-channel component, which split_channels() always constructs with channels_first=False (the ANTsImage.__init__ default). The explicit propagation code looks correct but is always a no-op.

ants.smooth_image's multi-component branch has the exact same loop-variable shadowing (for image in imagelist:), and on top of that didn't even attempt to pass channels_first to merge_channels at all.

ANTsImage.__getitem__'s has_components branch had the same omission as smooth_image — no channels_first passed through.

Fix

• resample_image: capture channels_first = image.channels_first before the loop, and rename the loop variable so it can no longer shadow the function's own image parameter.
• smooth_image: same pattern — capture before the loop, rename the loop variable, and pass channels_first through to merge_channels (previously missing entirely).
• ANTsImage.__getitem__: pass channels_first=self.channels_first through to merge_channels.

Testing

Verified live against the real compiled ants package (pip install antspyx, patching in the fixed pure-Python files):

import ants
img1 = ants.image_read(ants.get_ants_data('r16'))
img2 = ants.image_read(ants.get_ants_data('r16'))
vecimg = ants.merge_channels([img1, img2], channels_first=True)

resampled = ants.resample_image(vecimg, (100, 100), True)
print(resampled.channels_first)  # True (was False before the fix)

smoothed = ants.smooth_image(vecimg, 1.0)
print(smoothed.channels_first)  # True (was False before the fix)

Worth noting: my first attempt at the smooth_image fix (just adding channels_first=image.channels_first to the existing merge_channels call, matching the pattern suggested when this was scoped out in #938's review) was still broken by the same loop-shadowing bug — caught by this live check, which is why the fix renames the loop variable there too rather than only adding the keyword argument.

I did not have a way to exercise ANTsImage.__getitem__'s has_components branch through the pip-installed wheel in the time available, but the change there is the same pattern (no loop-variable shadowing risk, since it's a list comprehension using self.channels_first directly, not a loop variable).

AI Usage Disclosure

This PR was put together with AI assistance (Claude Code). Every change was traced against the real source, verified live against the installed package as shown above, and reviewed by me before pushing.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has insufficient remaining quota. Try requesting a review at a lower effort level.