Add missing channel_wise parameter to RandScaleIntensityFixedMean by engmohamedsalah · Pull Request #8741 · Project-MONAI/MONAI

engmohamedsalah · 2026-02-14T20:28:48Z

Summary

Adds the channel_wise parameter to RandScaleIntensityFixedMean and RandScaleIntensityFixedMeand, which was documented in docstrings but never implemented.
When channel_wise=True, a separate random scale factor is generated per channel, and preserve_range/fixed_mean are applied per channel — following the existing pattern from RandScaleIntensity.
Fixes docstring indentation for channel_wise in both array and dictionary transforms.

Fixes #8363

Changes

`monai/transforms/intensity/array.py`

Added channel_wise parameter to RandScaleIntensityFixedMean.__init__
Updated randomize() to generate per-channel factors when channel_wise=True
Updated __call__ to apply per-channel scaling with individual random factors

`monai/transforms/intensity/dictionary.py`

Added channel_wise parameter to RandScaleIntensityFixedMeand.__init__
Updated __call__ to pass image data to randomize() (needed for channel count), following the pattern from RandScaleIntensityd

Tests

Added test_channel_wise and test_channel_wise_preserve_range to test_rand_scale_intensity_fixed_mean.py
Added test_channel_wise to test_rand_scale_intensity_fixed_meand.py

Test plan

All existing tests pass (no regressions)
New channel_wise tests pass for both array and dictionary transforms
Verified per-channel mean preservation with fixed_mean=True
Verified preserve_range clipping works per channel

coderabbitai · 2026-02-14T20:29:05Z

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Cache: Disabled due to data retention organization setting

Knowledge base: Disabled due to Reviews -> Disable Knowledge Base setting

📥 Commits

Reviewing files that changed from the base of the PR and between a1f8023 and f996a93.

📒 Files selected for processing (4)

monai/transforms/intensity/array.py
monai/transforms/intensity/dictionary.py
tests/transforms/test_rand_scale_intensity_fixed_mean.py
tests/transforms/test_rand_scale_intensity_fixed_meand.py

🚧 Files skipped from review as they are similar to previous changes (1)

tests/transforms/test_rand_scale_intensity_fixed_meand.py

📝 Walkthrough

Walkthrough

Adds channel-wise support to RandScaleIntensityFixedMean and its dictionary wrapper. The array transform receives a new channel_wise bool parameter, forwards it to the internal scaler, randomizes per-channel factors when enabled, and applies per-channel scaling by iterating channels, stacking results, and preserving dtype. The dictionary transform propagates channel_wise, adjusts randomization to reuse a key-specific factor, and includes an early tensor-conversion path for empty-key cases. Tests were added to validate per-channel scaling and preserve_range behavior.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~25 minutes

Additional notes

Implementation defaults channel_wise=False to preserve backward compatibility. Review should focus on correctness of per-channel factor generation, channel iteration/stacking, dtype/preserve_range handling, and the adjusted dictionary call paths.

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 27.27% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	Title accurately summarizes the main change: implementing the missing channel_wise parameter for RandScaleIntensityFixedMean.
Description check	✅ Passed	Description is comprehensive and well-structured, covering changes in multiple files, test additions, and linking to the issue. All key details are present.
Linked Issues check	✅ Passed	The PR fully addresses issue `#8363` by implementing the documented but missing channel_wise parameter with per-channel factor generation and per-channel application of preserve_range and fixed_mean.
Out of Scope Changes check	✅ Passed	All changes are directly scoped to implementing channel_wise support across array and dictionary transforms plus corresponding tests, with no extraneous modifications.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings (stacked PR)
📝 Generate docstrings (commit on current branch)

🧪 Generate unit tests (beta)

Create PR with unit tests
Post copyable unit tests in a comment

Tip

Issue Planner is now in beta. Read the docs and try it out! Share your feedback on Discord.

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

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

…oject-MONAI#8363) The channel_wise parameter was documented in the docstring but not actually accepted by RandScaleIntensityFixedMean or its dictionary variant. This adds the parameter with per-channel random factor generation and scaling, matching the existing pattern in RandScaleIntensity. Also fixes docstring indentation. Signed-off-by: Mohamed Salah <eng.mohamed.tawab@gmail.com>

…ean-8363

ericspod

Hi @engmohamedsalah I made a few comments below but it otherwise looks good to me with a few changes. Thanks!

ericspod · 2026-03-01T21:58:24Z

monai/transforms/intensity/array.py

        factors: Sequence[float] | float = 0,
        fixed_mean: bool = True,
        preserve_range: bool = False,
+        channel_wise: bool = False,


Please add new arguments to the end of the argument list so that positional arguments aren't affected. I realise the argument was meant to be here given the docstring ordering, but if users call this constructor with positional arguments this will break compatibility.

ericspod · 2026-03-01T22:06:17Z

monai/transforms/intensity/array.py

            return None
-        self.factor = self.R.uniform(low=self.factors[0], high=self.factors[1])
+        if self.channel_wise:
+            self.factor = [self.R.uniform(low=self.factors[0], high=self.factors[1]) for _ in range(data.shape[0])]  # type: ignore


Suggested change

self.factor = [self.R.uniform(low=self.factors[0], high=self.factors[1]) for _ in range(data.shape[0])] # type: ignore

self.factor = self.R.uniform(low=self.factors[0], high=self.factors[1], size=data.shape[:1]) # type: ignore

This might be slightly better.

ericspod · 2026-03-01T22:13:07Z

monai/transforms/intensity/array.py

+        if self.channel_wise:
+            out = []
+            for i, d in enumerate(img):
+                out_channel = ScaleIntensityFixedMean(
+                    factor=self.factor[i],  # type: ignore
+                    fixed_mean=self.fixed_mean,
+                    preserve_range=self.preserve_range,
+                    dtype=self.dtype,
+                )(d[None])[0]
+                out.append(out_channel)
+            ret: NdarrayOrTensor = torch.stack(out)
+            ret = convert_to_dst_type(ret, dst=img, dtype=self.dtype or img.dtype)[0]
+            return ret


Suggested change

if self.channel_wise:

out = []

for i, d in enumerate(img):

out_channel = ScaleIntensityFixedMean(

factor=self.factor[i], # type: ignore

fixed_mean=self.fixed_mean,

preserve_range=self.preserve_range,

dtype=self.dtype,

)(d[None])[0]

out.append(out_channel)

ret: NdarrayOrTensor = torch.stack(out)

ret = convert_to_dst_type(ret, dst=img, dtype=self.dtype or img.dtype)[0]

return ret

if self.channel_wise:

out = []

for i, d in enumerate(img):

scale_trans = ScaleIntensityFixedMean(

factor=float(self.factor[i]),

fixed_mean=self.fixed_mean,

preserve_range=self.preserve_range,

dtype=self.dtype,

)

out.append(scale_trans(d[None]))

ret: NdarrayOrTensor = torch.cat(out)

ret = convert_to_dst_type(ret, dst=img, dtype=self.dtype or img.dtype)[0]

return ret

I think this is a little more readable, the type ignore for factor may still be needed though.

ericspod · 2026-03-01T22:13:58Z

monai/transforms/intensity/dictionary.py

        factors: Sequence[float] | float,
        fixed_mean: bool = True,
        preserve_range: bool = False,
+        channel_wise: bool = False,


Same issue here, despite the docstring argument being incorrectly included in this order.

engmohamedsalah requested review from KumoLiu, Nic-Ma and ericspod as code owners February 14, 2026 20:28

engmohamedsalah force-pushed the fix/add-channel-wise-randscaleintensityfixedmean-8363 branch from 738a3f9 to da88dbe Compare February 23, 2026 14:22

engmohamedsalah force-pushed the fix/add-channel-wise-randscaleintensityfixedmean-8363 branch from da88dbe to f996a93 Compare February 23, 2026 15:43

engmohamedsalah added 2 commits February 25, 2026 14:32

Merge branch 'dev' into fix/add-channel-wise-randscaleintensityfixedm…

6516499

…ean-8363

Merge branch 'dev' into fix/add-channel-wise-randscaleintensityfixedm…

5918a07

…ean-8363

ericspod requested changes Mar 1, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add missing channel_wise parameter to RandScaleIntensityFixedMean#8741

Add missing channel_wise parameter to RandScaleIntensityFixedMean#8741
engmohamedsalah wants to merge 3 commits intoProject-MONAI:devfrom
engmohamedsalah:fix/add-channel-wise-randscaleintensityfixedmean-8363

engmohamedsalah commented Feb 14, 2026

Uh oh!

coderabbitai bot commented Feb 14, 2026 •

edited

Loading

Walkthrough

Estimated code review effort

Additional notes

❌ Failed checks (1 warning)

Uh oh!

ericspod left a comment

Uh oh!

ericspod Mar 1, 2026

Uh oh!

ericspod Mar 1, 2026

Uh oh!

ericspod Mar 1, 2026

Uh oh!

ericspod Mar 1, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

	self.factor = [self.R.uniform(low=self.factors[0], high=self.factors[1]) for _ in range(data.shape[0])] # type: ignore
	self.factor = self.R.uniform(low=self.factors[0], high=self.factors[1], size=data.shape[:1]) # type: ignore

Conversation

engmohamedsalah commented Feb 14, 2026

Summary

Changes

monai/transforms/intensity/array.py

monai/transforms/intensity/dictionary.py

Tests

Test plan

Uh oh!

coderabbitai bot commented Feb 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Walkthrough

Estimated code review effort

Additional notes

❌ Failed checks (1 warning)

Uh oh!

ericspod left a comment

Choose a reason for hiding this comment

Uh oh!

ericspod Mar 1, 2026

Choose a reason for hiding this comment

Uh oh!

ericspod Mar 1, 2026

Choose a reason for hiding this comment

Uh oh!

ericspod Mar 1, 2026

Choose a reason for hiding this comment

Uh oh!

ericspod Mar 1, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

`monai/transforms/intensity/array.py`

`monai/transforms/intensity/dictionary.py`

coderabbitai bot commented Feb 14, 2026 •

edited

Loading