Feature added: Get WSI at mpp

[NikolasSchmitz]

Fixes: #4980

Description

In this pull request, the feature to retrieve a whole slide image at a given mpp (microns per pixel) resolution was implemented for every WSIReader class in the function get_wsi_at_mpp.

While the implementations in the OpenslideWSIReader and CuCIMWSIReader classes were tested thoroughly, I could not find a suitable TIFF file for testing with the TiffFileWSIReader class.

For resizing, I have used PIL.Image.resize for Openslide and TiffFile, and cucim.sklearn.transform.resize for CuCIM. Originally, I used cv2.resize, but since the package isn't listed in requirements-dev.txt, I explored alternative solutions."

Types of changes

• Non-breaking change (fix or new feature that would not break existing functionality).
• Breaking change (fix or new feature that would cause existing functionality to change).
• New tests added to cover the changes.
• Integration tests passed locally by running ./runtests.sh -f -u --net --coverage.
• Quick tests passed locally by running ./runtests.sh --quick --unittests --disttests.
• In-line docstrings updated.
• Documentation updated, tested make html command in the docs/ folder.

Please let me know what you think and how I can improve this feature.

Best
Niko

(Last PR was closed, because I changed the branch name to include the ticket id)

[bhashemian]

@NikolasSchmitz, I resolved the conflicts and updated the PR. Let's focus on the functionality and make this PR ready for reviewing. We can take care of any other issue once the PR is ready and please feel free to reach out to me or the working group if you still have any questions. Thanks for taking on this feature.

[NikolasSchmitz]

Thank you @drbeh ! I now marked it as ready for review.

[ericspod]

@drbeh would you be able to review this? I think you're best qualified here. I made some minor comments about print and code duplication but otherwise it seems fine to me without having tested it. Thanks!

[JHancox]

Thanks for the effort @NikolasSchmitz. A welcome addition! I agree with @ericspod about the refactoring to remove duplication - that would be great.

[NikolasSchmitz]

Thank you for the feedback, @JHancox . Unfortunately, I couldn't join the meeting because I was on my way home from ICLR.
I will review the code and provide my updates as soon as possible.

[ericspod]

Hi @NikolasSchmitz thanks for getting back to this. There's a few tests which still aren't passing, a few say just that a timeout occurred so it might be internal but the code formatting test did find something. As for DCO, clicking on that brings up advice on how to resolve, specifically doing an empty remedial commit. If that doesn't work we will set the DCO to pass just before merging. Please have a look at the tests and see if we can get that through.

[ericspod]

We also are putting more test data on Huggingface now so you can propose to upload your image to here https://huggingface.co/datasets/MONAI/testing_data if you like. We're planning to adjust our tests to pull from this dataset.

[ericspod]

Your new tiff file is added to the hf dataset now, so you need to change setUpModule in test_wsireader.py to be:

@skipUnless(has_cucim or has_osl or has_tiff, "Requires cucim, openslide, or tifffile!")
def setUpModule():
    download_url_or_skip_test(
        testing_data_config("images", WSI_GENERIC_TIFF_KEY, "url"),
        WSI_GENERIC_TIFF_PATH,
        hash_type=testing_data_config("images", WSI_GENERIC_TIFF_KEY, "hash_type"),
        hash_val=testing_data_config("images", WSI_GENERIC_TIFF_KEY, "hash_val"),
    )
    download_url_or_skip_test(
        testing_data_config("images", WSI_GENERIC_TIFF_CORRECT_MPP_KEY, "url"),
        WSI_GENERIC_TIFF_CORRECT_MPP_PATH,
        hash_type=testing_data_config("images", WSI_GENERIC_TIFF_CORRECT_MPP_KEY, "hash_type"),
        hash_val=testing_data_config("images", WSI_GENERIC_TIFF_CORRECT_MPP_KEY, "hash_val"),
    )
    download_url_or_skip_test(
        testing_data_config("images", WSI_APERIO_SVS_KEY, "url"),
        WSI_APERIO_SVS_PATH,
        hash_type=testing_data_config("images", WSI_APERIO_SVS_KEY, "hash_type"),
        hash_val=testing_data_config("images", WSI_APERIO_SVS_KEY, "hash_val"),
    )

Add the following to tests/testing_data/data_config.json just after the first item in the images section:

        "wsi_generic_tiff_corrected": {
            "url": "https://huggingface.co/datasets/MONAI/testing_data/resolve/main/CMU-1_correct_mpp.tiff",
            "hash_type": "sha256",
            "hash_val": "65306e3f8f7f5282d19d942dadc525cd06a80d5fd8268053939751365226c65f"
        },

As you can see, our process for downloading data needs to be revamped to improved this process! Thanks.

[bhashemian]

@NikolasSchmitz thanks for your contribution. Please let me know if you need help fixing the DCO issues.

[bhashemian]

@coderabbitai full review

[coderabbitai]

✅ Actions performed

Full review triggered.

[coderabbitai]

Actionable comments posted: 3

♻️ Duplicate comments (1)

monai/data/wsi_reader.py (1)

663-681: Validate mpp input early.

Reject non-tuples, wrong length, or non-positive values to avoid undefined behavior downstream.

 def get_wsi_at_mpp(self, wsi, mpp: tuple, atol: float = 0.00, rtol: float = 0.05) -> np.ndarray:
@@
-        return self.reader.get_wsi_at_mpp(wsi, mpp, atol, rtol)
+        # basic validation
+        if not isinstance(mpp, tuple) or len(mpp) != 2:
+            raise ValueError("mpp must be a tuple (x, y).")
+        if any((not isinstance(v, (int, float))) or v <= 0 for v in mpp):
+            raise ValueError("mpp values must be positive numbers.")
+        return self.reader.get_wsi_at_mpp(wsi, mpp, atol, rtol)

🧹 Nitpick comments (8)

tests/utils/enums/test_wsireader.py (2)

479-505: Strengthen assertions for cross-backend robustness.

Exact H×W can vary a few pixels due to rounding/interp differences. Also assert:

• dtype and channel count per backend,
• monotonicity of area vs requested MPP,
• that returned MPP (via metadata or helper) is within tolerance.

Consider computing expected size from level-0 dims and requested mpp rather than hard-coding tuples to reduce brittleness.

---

467-478: Remove commented-out test or re-enable behind a marker.

Dead/commented code adds noise. Delete or gate with a feature flag/xfail.

monai/data/wsi_reader.py (5)

843-846: Don’t pass hard-coded tolerances to _find_closest_level; avoid unintended behavior.

The 0,5 literals silently decouple user-provided tolerances. Either:

• use the provided (atol, rtol), or
• bypass tolerance checking entirely when only the index of the closest level is needed.

Apply one of the following:

Option A (simple):

-        closest_lvl = self._find_closest_level("mpp", mpp, mpp_list, 0, 5)
+        closest_lvl = self._find_closest_level("mpp", mpp, mpp_list, atol, rtol)

Option B (preferred: no early raising):

-        closest_lvl = self._find_closest_level("mpp", mpp, mpp_list, 0, 5)
+        # Pick closest level without tolerance checks; actual tolerances are evaluated below.
+        closest_lvl = min(range(len(mpp_list)),
+                          key=lambda i: abs(mpp_list[i][0]-mpp[0]) + abs(mpp_list[i][1]-mpp[1]))

Also applies to: 1108-1111, 1345-1348

---

949-975: Normalize dimension handling to reduce fragility.

closest_lvl_dim is backend-native (often (width, height)); _compute_mpp_target_res expects (height, width), leading to implicit swaps. Unify by passing self.get_size(wsi, closest_lvl) to _compute_mpp_target_res, and use explicit width/height when calling backend resizers.

Example for CuCIM (apply similarly to OpenSlide/TiffFile as appropriate):

-        closest_lvl_dim = wsi.resolutions["level_dimensions"][closest_lvl]
-        target_res_x, target_res_y = self._compute_mpp_target_res(closest_lvl, closest_lvl_dim, mpp_list, user_mpp)
-        wsi_arr = cp.array(wsi.read_region((0, 0), level=closest_lvl, size=closest_lvl_dim, num_workers=self.num_workers))
+        # Consistent dims: (h, w)
+        h, w = self.get_size(wsi, closest_lvl)
+        target_w, target_h = self._compute_mpp_target_res(closest_lvl, (h, w), mpp_list, user_mpp)
+        # cuCIM read_region expects (w, h)
+        wsi_arr = cp.array(wsi.read_region((0, 0), level=closest_lvl, size=(w, h), num_workers=self.num_workers))
-        closest_lvl_wsi = cucim_resize(
-            wsi_arr,
-            (target_res_x, target_res_y),
+        # cucim.resize expects (rows, cols) == (h, w)
+        closest_lvl_wsi = cucim_resize(
+            wsi_arr,
+            (target_h, target_w),
             order=1,
             preserve_range=True,
             anti_aliasing=False).astype(cp.uint8)

For OpenSlide’s PIL.resize, pass (target_w, target_h).

Also applies to: 1200-1221, 1442-1462

---

1314-1324: Normalize ResolutionUnit and tighten error handling.

Lowercase the unit to match ConvertUnits expectations; use shorter messages (TRY003).

-            convert_to_micron = ConvertUnits(unit, "micrometer")
+            convert_to_micron = ConvertUnits(str(unit).lower(), "micrometer")
@@
-            if xres[0] and yres[0]:
+            if xres[0] and yres[0]:
                 return convert_to_micron(yres[1] / yres[0]), convert_to_micron(xres[1] / xres[0])
             else:
-                raise ValueError("The `XResolution` and/or `YResolution` property of the image is zero, "
-                                 "which is needed to obtain `mpp` for this file. Please use `level` instead.")
+                raise ValueError("Invalid TIFF X/YResolution (zero). Cannot compute mpp; use `level` instead.")

---

960-974: Consider anti-aliasing when downscaling.

Downscaling without anti_aliasing can introduce artifacts. Enable it when target size is smaller than source.

-        closest_lvl_wsi = cucim_resize(
+        downscale = (target_h < h) and (target_w < w)
+        closest_lvl_wsi = cucim_resize(
             wsi_arr,
             (target_h, target_w),
             order=1,
             preserve_range=True,
-            anti_aliasing=False).astype(cp.uint8)
+            anti_aliasing=downscale).astype(cp.uint8)

---

1099-1105: Docstring: clarify tolerance interplay and behavior on miss.

Explicitly state: closest level is selected ignoring tolerances; tolerances only decide “return as-is vs resize”. Document behavior when requested MPP is finer than level 0 (error vs upsample).

Also applies to: 1336-1342, 833-839

tests/testing_data/data_config.json (1)

8-12: Consider pinning to a specific commit for reproducibility.

Hash verified ✓ and artifact accessible. However, resolve/main remains mutable; prefer pinning to a commit hash (e.g., resolve/abc1234) for CI stability. The SHA256 guard mitigates risk but doesn't prevent future mutations.

📜 Review details

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 69f3dd2 and 1c1ecd0.

📒 Files selected for processing (4)

• monai/data/wsi_reader.py (9 hunks)
• monai/transforms/regularization/array.py (0 hunks)
• tests/testing_data/data_config.json (1 hunks)
• tests/utils/enums/test_wsireader.py (4 hunks)

💤 Files with no reviewable changes (1)

• monai/transforms/regularization/array.py

🧰 Additional context used

📓 Path-based instructions (1)

**/*.py

⚙️ CodeRabbit configuration file

Review the Python code for quality and correctness. Ensure variable names adhere to PEP8 style guides, are sensible and informative in regards to their function, though permitting simple names for loop and comprehension variables. Ensure routine names are meaningful in regards to their function and use verbs, adjectives, and nouns in a semantically appropriate way. Docstrings should be present for all definition which describe each variable, return value, and raised exception in the appropriate section of the Google-style of docstrings. Examine code for logical error or inconsistencies, and suggest what may be changed to addressed these. Suggest any enhancements for code improving efficiency, maintainability, comprehensibility, and correctness. Ensure new or modified definitions will be covered by existing or new unit tests.

Files:

• monai/data/wsi_reader.py
• tests/utils/enums/test_wsireader.py

🪛 Ruff (0.14.1)

monai/data/wsi_reader.py

1322-1323: Avoid specifying long messages outside the exception class

(TRY003)

---

1324-1324: Avoid specifying long messages outside the exception class

(TRY003)

⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (4)

• GitHub Check: quick-py3 (windows-latest)
• GitHub Check: quick-py3 (macOS-latest)
• GitHub Check: packaging
• GitHub Check: quick-py3 (ubuntu-latest)

[coderabbitai]

🛠️ Refactor suggestion | 🟠 Major

Add a “too-fine MPP” test to catch boundary behavior.

Add a case where requested mpp is smaller than level-0 (i.e., finer than native). Current backend code decrements level and can underflow; a test will guard it.

I can add a parameterized TEST_CASE__MPP_TOO_FINE and expected exception once backend behavior is decided (error vs upsample).

Also applies to: 286-309

[bhashemian]

@NikolasSchmitz please let me know if I can help to finalize and merge this PR. Thanks

[NikolasSchmitz]

Hey Bruce, I will finalize this PR in the next days and will let you know whether I need help. Thanks!

[bhashemian]

/build

[NikolasSchmitz]

Hi @bhashemian, I’ve just pushed my current progress which includes two additions suggested by coderabbitai.
I could use a hand to sort out the DCO issues, or some hint on how I can solve this easily.
Also, I’ve noticed issues with the output whenever the user-specified MPP is below the image’s native MPP resolution. I’ll try to sort those out.

Thanks!

[coderabbitai]

Actionable comments posted: 9

♻️ Duplicate comments (5)

monai/data/wsi_reader.py (5)

843-843: Fix hardcoded tolerance values.

Line uses hardcoded (0, 5) instead of the atol and rtol parameters passed to the function. This bug was flagged in previous reviews.

Apply this diff:

-        closest_lvl = self._find_closest_level("mpp", mpp, mpp_list, 0, 5)
+        closest_lvl = self._find_closest_level("mpp", mpp, mpp_list, atol, rtol)

---

1110-1110: Fix hardcoded tolerance values.

Line uses hardcoded (0, 5) instead of the atol and rtol parameters.

Apply this diff:

-        closest_lvl = self._find_closest_level("mpp", mpp, mpp_list, 0, 5)
+        closest_lvl = self._find_closest_level("mpp", mpp, mpp_list, atol, rtol)

---

1124-1127: Guard against level underflow.

Decrementing closest_lvl without checking if it's 0 can cause underflow to -1.

Apply this diff:

         else:
             # If both checks fail, increase resolution (i.e., decrement level) and then downsample it.
+            if closest_lvl == 0:
+                closest_lvl_wsi = self._resize_to_mpp_res(wsi, closest_lvl, mpp_list, mpp)
+            else:
-            closest_lvl = closest_lvl - 1
-            closest_lvl_wsi = self._resize_to_mpp_res(wsi, closest_lvl, mpp_list, mpp)
+                closest_lvl = closest_lvl - 1
+                closest_lvl_wsi = self._resize_to_mpp_res(wsi, closest_lvl, mpp_list, mpp)

---

1347-1347: Fix hardcoded tolerance values.

Line uses hardcoded (0, 5) instead of the atol and rtol parameters.

Apply this diff:

-        closest_lvl = self._find_closest_level("mpp", mpp, mpp_list, 0, 5)
+        closest_lvl = self._find_closest_level("mpp", mpp, mpp_list, atol, rtol)

---

1359-1362: Guard against level underflow.

Decrementing closest_lvl without checking if it's 0 can cause underflow to -1.

Apply this diff:

         else:
             # If both checks fail, increase resolution (i.e., decrement level) and then downsample it.
+            if closest_lvl == 0:
+                closest_lvl_wsi = self._resize_to_mpp_res(wsi, closest_lvl, mpp_list, mpp)
+            else:
-            closest_lvl = closest_lvl - 1
-            closest_lvl_wsi = self._resize_to_mpp_res(wsi, closest_lvl, mpp_list, mpp)
+                closest_lvl = closest_lvl - 1
+                closest_lvl_wsi = self._resize_to_mpp_res(wsi, closest_lvl, mpp_list, mpp)

📜 Review details

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 1c1ecd0 and 165af63.

📒 Files selected for processing (1)

• monai/data/wsi_reader.py (9 hunks)

🧰 Additional context used

📓 Path-based instructions (1)

**/*.py

⚙️ CodeRabbit configuration file

Review the Python code for quality and correctness. Ensure variable names adhere to PEP8 style guides, are sensible and informative in regards to their function, though permitting simple names for loop and comprehension variables. Ensure routine names are meaningful in regards to their function and use verbs, adjectives, and nouns in a semantically appropriate way. Docstrings should be present for all definition which describe each variable, return value, and raised exception in the appropriate section of the Google-style of docstrings. Examine code for logical error or inconsistencies, and suggest what may be changed to addressed these. Suggest any enhancements for code improving efficiency, maintainability, comprehensibility, and correctness. Ensure new or modified definitions will be covered by existing or new unit tests.

Files:

• monai/data/wsi_reader.py

🪛 Ruff (0.14.4)

monai/data/wsi_reader.py

1324-1325: Avoid specifying long messages outside the exception class

(TRY003)

---

1326-1326: Avoid specifying long messages outside the exception class

(TRY003)

⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (19)

• GitHub Check: packaging
• GitHub Check: build-docs
• GitHub Check: quick-py3 (windows-latest)
• GitHub Check: flake8-py3 (pytype)
• GitHub Check: quick-py3 (macOS-latest)
• GitHub Check: quick-py3 (ubuntu-latest)
• GitHub Check: flake8-py3 (codeformat)
• GitHub Check: flake8-py3 (mypy)
• GitHub Check: min-dep-pytorch (2.6.0)
• GitHub Check: min-dep-os (windows-latest)
• GitHub Check: min-dep-pytorch (2.8.0)
• GitHub Check: min-dep-os (ubuntu-latest)
• GitHub Check: min-dep-pytorch (2.7.1)
• GitHub Check: min-dep-pytorch (2.5.1)
• GitHub Check: min-dep-py3 (3.9)
• GitHub Check: min-dep-py3 (3.11)
• GitHub Check: min-dep-py3 (3.12)
• GitHub Check: min-dep-py3 (3.10)
• GitHub Check: min-dep-os (macOS-latest)

🔇 Additional comments (1)

monai/data/wsi_reader.py (1)

1317-1326: Validation logic for zero resolution is correct.

The check prevents division-by-zero when X/YResolution numerators are zero. Static analysis flagged the long exception message (TRY003), but this is acceptable for clarity.

[coderabbitai]

Actionable comments posted: 1

♻️ Duplicate comments (5)

monai/data/wsi_reader.py (5)

843-843: Fix hardcoded tolerance values.

Uses hardcoded (0, 5) instead of the function's atol and rtol parameters. This bug was flagged in past reviews but remains unfixed.

Apply this diff:

-        closest_lvl = self._find_closest_level("mpp", mpp, mpp_list, 0, 5)
+        closest_lvl = self._find_closest_level("mpp", mpp, mpp_list, atol, rtol)

---

1110-1110: Fix hardcoded tolerance values.

Uses hardcoded (0, 5) instead of the function's atol and rtol parameters. Same bug as in CuCIMWSIReader.

Apply this diff:

-        closest_lvl = self._find_closest_level("mpp", mpp, mpp_list, 0, 5)
+        closest_lvl = self._find_closest_level("mpp", mpp, mpp_list, atol, rtol)

---

1359-1362: Guard against level underflow.

If closest_lvl is 0, line 1361 decrements it to -1, causing negative indexing.

Apply this diff:

         else:
             # If both checks fail, increase resolution (i.e., decrement level) and then downsample it.
-            closest_lvl = closest_lvl - 1
-            closest_lvl_wsi = self._resize_to_mpp_res(wsi, closest_lvl, mpp_list, mpp)
+            if closest_lvl == 0:
+                closest_lvl_wsi = self._resize_to_mpp_res(wsi, closest_lvl, mpp_list, mpp)
+            else:
+                closest_lvl = closest_lvl - 1
+                closest_lvl_wsi = self._resize_to_mpp_res(wsi, closest_lvl, mpp_list, mpp)

---

1347-1347: Fix hardcoded tolerance values.

Uses hardcoded (0, 5) instead of the function's atol and rtol parameters. Same bug as in other backends.

Apply this diff:

-        closest_lvl = self._find_closest_level("mpp", mpp, mpp_list, 0, 5)
+        closest_lvl = self._find_closest_level("mpp", mpp, mpp_list, atol, rtol)

---

1124-1127: Guard against level underflow.

If closest_lvl is 0, line 1126 decrements it to -1, causing negative indexing into the level list.

Apply this diff:

         else:
             # If both checks fail, increase resolution (i.e., decrement level) and then downsample it.
-            closest_lvl = closest_lvl - 1
-            closest_lvl_wsi = self._resize_to_mpp_res(wsi, closest_lvl, mpp_list, mpp)
+            if closest_lvl == 0:
+                closest_lvl_wsi = self._resize_to_mpp_res(wsi, closest_lvl, mpp_list, mpp)
+            else:
+                closest_lvl = closest_lvl - 1
+                closest_lvl_wsi = self._resize_to_mpp_res(wsi, closest_lvl, mpp_list, mpp)

🧹 Nitpick comments (3)

monai/data/wsi_reader.py (3)

434-455: Document return type in docstring.

The method returns tuple[int, int] (target width, height) but this isn't specified in the docstring.

Add to docstring:

         Args:
             closest_lvl: Whole slide image level closest to user-provided MPP resolution.
             closest_lvl_dim: Dimensions (height, width) of the image at the closest level.
             mpp_list: List of MPP values for all levels of the whole slide image.
             mpp: The MPP resolution at which the whole slide image representation should be extracted.
+
+        Returns:
+            tuple[int, int]: Target dimensions (width, height) for the resized image.

---

457-489: Document return type in docstring.

The method returns tuple[bool, bool] but this isn't specified in the docstring.

Add to docstring:

         Args:
             closest_lvl: Whole slide image level closest to user-provided MPP resolution.
             mpp_list: List of MPP values for all levels of the whole slide image.
             mpp: The MPP resolution at which the whole slide image representation should be extracted.
             atol: Absolute tolerance for MPP comparison.
             rtol: Relative tolerance for MPP comparison.
+
+        Returns:
+            tuple[bool, bool]: (is_within_tolerance, closest_level_is_bigger)

---

951-977: Minor formatting: add space before method call.

Line 975 missing space before .astype.

Apply this diff:

             (target_res_x, target_res_y),
             order=1,
             preserve_range=True,
-            anti_aliasing=False).astype(cp.uint8)
+            anti_aliasing=False
+        ).astype(cp.uint8)

📜 Review details

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 1c1ecd0 and 165af63.

📒 Files selected for processing (1)

• monai/data/wsi_reader.py (9 hunks)

🧰 Additional context used

📓 Path-based instructions (1)

**/*.py

⚙️ CodeRabbit configuration file

Review the Python code for quality and correctness. Ensure variable names adhere to PEP8 style guides, are sensible and informative in regards to their function, though permitting simple names for loop and comprehension variables. Ensure routine names are meaningful in regards to their function and use verbs, adjectives, and nouns in a semantically appropriate way. Docstrings should be present for all definition which describe each variable, return value, and raised exception in the appropriate section of the Google-style of docstrings. Examine code for logical error or inconsistencies, and suggest what may be changed to addressed these. Suggest any enhancements for code improving efficiency, maintainability, comprehensibility, and correctness. Ensure new or modified definitions will be covered by existing or new unit tests.

Files:

• monai/data/wsi_reader.py

🪛 Ruff (0.14.4)

monai/data/wsi_reader.py

1324-1325: Avoid specifying long messages outside the exception class

(TRY003)

---

1326-1326: Avoid specifying long messages outside the exception class

(TRY003)

⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (3)

• GitHub Check: quick-py3 (windows-latest)
• GitHub Check: quick-py3 (macOS-latest)
• GitHub Check: quick-py3 (ubuntu-latest)

🔇 Additional comments (3)

monai/data/wsi_reader.py (3)

1202-1222: LGTM.

Resizing logic is correct and uses appropriate PIL interpolation.

---

1317-1326: LGTM - division-by-zero guard added.

The validation correctly prevents division by zero when TIFF resolution tags contain zero numerators. The static analysis hint (TRY003) about exception message length is pedantic and can be ignored here.

---

1444-1464: LGTM.

Resizing logic is correct and consistent with OpenSlide implementation.

[coderabbitai]

Actionable comments posted: 1

🧹 Nitpick comments (1)

monai/data/wsi_reader.py (1)

460-496: Fix return type annotation.

_compute_mpp_tolerances actually returns a pair of booleans; the annotation -> bool is misleading and confuses type checkers. Update it to tuple[bool, bool] to match the real contract.

📜 Review details

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 165af63 and 4460f45.

📒 Files selected for processing (1)

• monai/data/wsi_reader.py (9 hunks)

🧰 Additional context used

📓 Path-based instructions (1)

**/*.py

⚙️ CodeRabbit configuration file

Review the Python code for quality and correctness. Ensure variable names adhere to PEP8 style guides, are sensible and informative in regards to their function, though permitting simple names for loop and comprehension variables. Ensure routine names are meaningful in regards to their function and use verbs, adjectives, and nouns in a semantically appropriate way. Docstrings should be present for all definition which describe each variable, return value, and raised exception in the appropriate section of the Google-style of docstrings. Examine code for logical error or inconsistencies, and suggest what may be changed to addressed these. Suggest any enhancements for code improving efficiency, maintainability, comprehensibility, and correctness. Ensure new or modified definitions will be covered by existing or new unit tests.

Files:

• monai/data/wsi_reader.py

🪛 Ruff (0.14.4)

monai/data/wsi_reader.py

1350-1351: Avoid specifying long messages outside the exception class

(TRY003)

---

1352-1352: Avoid specifying long messages outside the exception class

(TRY003)

⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (19)

• GitHub Check: min-dep-py3 (3.11)
• GitHub Check: min-dep-pytorch (2.8.0)
• GitHub Check: min-dep-py3 (3.10)
• GitHub Check: min-dep-py3 (3.9)
• GitHub Check: min-dep-pytorch (2.6.0)
• GitHub Check: min-dep-os (ubuntu-latest)
• GitHub Check: min-dep-pytorch (2.7.1)
• GitHub Check: min-dep-pytorch (2.5.1)
• GitHub Check: min-dep-py3 (3.12)
• GitHub Check: min-dep-os (windows-latest)
• GitHub Check: min-dep-os (macOS-latest)
• GitHub Check: flake8-py3 (pytype)
• GitHub Check: packaging
• GitHub Check: quick-py3 (windows-latest)
• GitHub Check: build-docs
• GitHub Check: quick-py3 (ubuntu-latest)
• GitHub Check: flake8-py3 (mypy)
• GitHub Check: quick-py3 (macOS-latest)
• GitHub Check: flake8-py3 (codeformat)

[coderabbitai]

⚠️ Potential issue | 🟠 Major

Handle scalar mpp inputs.

Passing a single float (supported elsewhere in this API surface) now breaks because _find_closest_level calls len(value) on the scalar and raises TypeError. Normalize mpp with ensure_tuple_rep(mpp, 2) before using it here, and mirror the guard in the OpenSlide and TiffFile implementations so direct backend use stays compatible with existing call patterns.

Apply this diff (and replicate in the other backends):

         cp, _ = optional_import("cupy")
 
+        mpp = ensure_tuple_rep(mpp, 2)
+
         mpp_list = [self.get_mpp(wsi, lvl) for lvl in range(wsi.resolutions["level_count"])]

🤖 Prompt for AI Agents

In monai/data/wsi_reader.py around lines 833 to 879, the function get_wsi_at_mpp
fails when mpp is passed as a scalar because _find_closest_level expects a tuple
and calls len(value); normalize mpp at the start of the function using
ensure_tuple_rep(mpp, 2) (same guard used in OpenSlide/TiffFile backends) so
subsequent calls (mpp_list computation and _find_closest_level) always receive a
2-tuple; apply the same normalization to the other backend implementations to
keep direct backend usage compatible with existing call patterns.