appease the formatter

gerlichlab · Mar 15, 2024 · 352e176 · 352e176
1 parent 536d0ad
commit 352e176
Show file tree

Hide file tree

Showing 5 changed files with 164 additions and 82 deletions.
diff --git a/noxfile.py b/noxfile.py
@@ -48,11 +48,13 @@ def install_groups(
         # See: https://python-poetry.org/blog/announcing-poetry-1.7.0/
         argv = [
             "poetry",
-            "config", 
+            "config",
             "warnings.export",
             "false",
         ]
-        session.debug(f"Running command to silence warning since in Nix shell we install the patch: {' '.join(argv)}")
+        session.debug(
+            f"Running command to silence warning since in Nix shell we install the patch: {' '.join(argv)}"
+        )
         session.run_always(*argv, external=True)
         session.log(f"Will generate requirements hashfile: {hashfile}")
         requirements_txt.parent.mkdir(parents=True, exist_ok=True)

diff --git a/spotfishing/deprecated.py b/spotfishing/deprecated.py
@@ -8,10 +8,13 @@
 from skimage.morphology import ball, remove_small_objects, white_tophat
 from skimage.segmentation import expand_labels
 
-
 Numeric = Union[int, float]
 
-CENTROID_COLUMNS_REMAPPING = {'centroid_weighted-0': 'zc', 'centroid_weighted-1': 'yc', 'centroid_weighted-2': 'xc'}
+CENTROID_COLUMNS_REMAPPING = {
+    "centroid_weighted-0": "zc",
+    "centroid_weighted-1": "yc",
+    "centroid_weighted-2": "xc",
+}
 
 
 def detect_spots_dog_old(input_img, spot_threshold: Numeric, expand_px: int = 10):
@@ -24,16 +27,16 @@ def detect_spots_dog_old(input_img, spot_threshold: Numeric, expand_px: int = 10
     spot_threshold : NumberLike
         Threshold to use for spots
     expand_px : int
-        Number of pixels by which to expand contiguous subregion, 
+        Number of pixels by which to expand contiguous subregion,
         up to point of overlap with neighboring subregion of image
-        
+
     Returns
     -------
-    pd.DataFrame, np.ndarray, np.ndarray: 
-        The centroids and roi_IDs of the spots found, 
-        the image used for spot detection, and 
-        numpy array with only sufficiently large regions 
-        retained (bigger than threshold number of pixels), 
+    pd.DataFrame, np.ndarray, np.ndarray:
+        The centroids and roi_IDs of the spots found,
+        the image used for spot detection, and
+        numpy array with only sufficiently large regions
+        retained (bigger than threshold number of pixels),
         and dilated by expansion amount (possibly)
     """
     # TODO: Do not use hard-coded sigma values (second parameter to gaussian(...)).
@@ -45,13 +48,19 @@ def detect_spots_dog_old(input_img, spot_threshold: Numeric, expand_px: int = 10
     img = (img - np.mean(img)) / np.std(img)
     labels, _ = ndi.label(img > spot_threshold)
     labels = expand_labels(labels, expand_px)
-    
+
     # Make a DataFrame with the ROI info.
-    spot_props = _reindex_to_roi_id(pd.DataFrame(regionprops_table(
-        label_image=labels, 
-        intensity_image=input_img, 
-        properties=('label', 'centroid_weighted', 'intensity_mean')
-        )).drop(['label'], axis=1).rename(columns=CENTROID_COLUMNS_REMAPPING))
+    spot_props = _reindex_to_roi_id(
+        pd.DataFrame(
+            regionprops_table(
+                label_image=labels,
+                intensity_image=input_img,
+                properties=("label", "centroid_weighted", "intensity_mean"),
+            )
+        )
+        .drop(["label"], axis=1)
+        .rename(columns=CENTROID_COLUMNS_REMAPPING)
+    )
 
     return spot_props, img, labels
 
@@ -66,16 +75,16 @@ def detect_spots_int_old(input_img, spot_threshold: Numeric, expand_px: int = 1)
     spot_threshold : NumberLike
         Threshold to use for spots
     expand_px : int
-        Number of pixels by which to expand contiguous subregion, 
+        Number of pixels by which to expand contiguous subregion,
         up to point of overlap with neighboring subregion of image
-        
+
     Returns
     -------
-    pd.DataFrame, np.ndarray, np.ndarray: 
-        The centroids and roi_IDs of the spots found, 
-        the image used for spot detection, and 
-        numpy array with only sufficiently large regions 
-        retained (bigger than threshold number of pixels), 
+    pd.DataFrame, np.ndarray, np.ndarray:
+        The centroids and roi_IDs of the spots found,
+        the image used for spot detection, and
+        numpy array with only sufficiently large regions
+        retained (bigger than threshold number of pixels),
         and dilated by expansion amount (possibly)
     """
     # TODO: enforce that output column names don't vary with code path walked.
@@ -85,33 +94,59 @@ def detect_spots_int_old(input_img, spot_threshold: Numeric, expand_px: int = 1)
     binary = ndi.binary_fill_holes(binary)
     struct = ndi.generate_binary_structure(input_img.ndim, 2)
     labels, n_obj = ndi.label(binary, structure=struct)
-    if n_obj > 1: # Do not need this with area filtering below
+    if n_obj > 1:  # Do not need this with area filtering below
         labels = remove_small_objects(labels, min_size=5)
     if expand_px > 0:
         labels = expand_labels(labels, expand_px)
-    if np.all(labels == 0): # No substructures (ROIs) exist after filtering.
-        spot_props = pd.DataFrame(columns=['label', 'z_min', 'y_min', 'x_min', 'z_max', 'y_max', 'x_max', 'area', 'zc', 'yc', 'xc', 'intensity_mean'])
+    if np.all(labels == 0):  # No substructures (ROIs) exist after filtering.
+        spot_props = pd.DataFrame(
+            columns=[
+                "label",
+                "z_min",
+                "y_min",
+                "x_min",
+                "z_max",
+                "y_max",
+                "x_max",
+                "area",
+                "zc",
+                "yc",
+                "xc",
+                "intensity_mean",
+            ]
+        )
     else:
         spot_props = _reindex_to_roi_id(
-            pd.DataFrame(regionprops_table(
-                labels, input_img, properties=('label', 'bbox', 'area', 'centroid_weighted', 'intensity_mean')
-                )).rename(
-                    columns={**CENTROID_COLUMNS_REMAPPING,
-                        'bbox-0': 'z_min',
-                        'bbox-1': 'y_min',
-                        'bbox-2': 'x_min',
-                        'bbox-3': 'z_max',
-                        'bbox-4': 'y_max',
-                        'bbox-5': 'x_max'
-                        }
-                    )
+            pd.DataFrame(
+                regionprops_table(
+                    labels,
+                    input_img,
+                    properties=(
+                        "label",
+                        "bbox",
+                        "area",
+                        "centroid_weighted",
+                        "intensity_mean",
+                    ),
+                )
+            ).rename(
+                columns={
+                    **CENTROID_COLUMNS_REMAPPING,
+                    "bbox-0": "z_min",
+                    "bbox-1": "y_min",
+                    "bbox-2": "x_min",
+                    "bbox-3": "z_max",
+                    "bbox-4": "y_max",
+                    "bbox-5": "x_max",
+                }
             )
+        )
 
     return spot_props, input_img, labels
 
 
 def _index_as_roi_id(props_table: pd.DataFrame) -> pd.DataFrame:
-    return props_table.rename(columns={'index': 'roi_id'})
+    return props_table.rename(columns={"index": "roi_id"})
 
 
 def _reindex_to_roi_id(props_table: pd.DataFrame) -> pd.DataFrame:

diff --git a/spotfishing/detectors.py b/spotfishing/detectors.py
@@ -9,8 +9,8 @@
 from skimage.measure import regionprops_table
 from skimage.morphology import ball, remove_small_objects, white_tophat
 from skimage.segmentation import expand_labels
-from .exceptions import DimensionalityError
 
+from .exceptions import DimensionalityError
 
 __author__ = "Vince Reuter"
 __credits__ = ["Vince Reuter", "Kai Sandoval Beckwith"]
@@ -19,10 +19,16 @@
 
 Numeric = Union[int, float]
 
-CENTROID_COLUMNS_REMAPPING = {'centroid_weighted-0': 'zc', 'centroid_weighted-1': 'yc', 'centroid_weighted-2': 'xc'}
+CENTROID_COLUMNS_REMAPPING = {
+    "centroid_weighted-0": "zc",
+    "centroid_weighted-1": "yc",
+    "centroid_weighted-2": "xc",
+}
 
 
-def detect_spots_dog(*, input_image, spot_threshold: Numeric, expand_px: Optional[Numeric]):
+def detect_spots_dog(
+    *, input_image, spot_threshold: Numeric, expand_px: Optional[Numeric]
+):
     """Spot detection by difference of Gaussians filter
 
     Arguments
@@ -32,26 +38,30 @@ def detect_spots_dog(*, input_image, spot_threshold: Numeric, expand_px: Optiona
     spot_threshold : int or float
         Minimum peak value after the DoG transformation to call a peak/spot
     expand_px : float or int or NoneType
-        Number of pixels by which to expand contiguous subregion, 
+        Number of pixels by which to expand contiguous subregion,
         up to point of overlap with neighboring subregion of image
-    
+
     Returns
     -------
-    pd.DataFrame, np.ndarray, np.ndarray: 
-        The centroids and roi_IDs of the spots found, 
-        the image used for spot detection, and 
-        numpy array with only sufficiently large regions 
-        retained (bigger than threshold number of pixels), 
+    pd.DataFrame, np.ndarray, np.ndarray:
+        The centroids and roi_IDs of the spots found,
+        the image used for spot detection, and
+        numpy array with only sufficiently large regions
+        retained (bigger than threshold number of pixels),
         and dilated by expansion amount (possibly)
     """
     _check_input_image(input_image)
     img = _preprocess_for_difference_of_gaussians(input_image)
     labels, _ = ndi.label(img > spot_threshold)
-    spot_props, labels = _build_props_table(labels=labels, input_image=input_image, expand_px=expand_px)
+    spot_props, labels = _build_props_table(
+        labels=labels, input_image=input_image, expand_px=expand_px
+    )
     return spot_props, img, labels
 
 
-def detect_spots_int(*, input_image, spot_threshold: Numeric, expand_px: Optional[Numeric]):
+def detect_spots_int(
+    *, input_image, spot_threshold: Numeric, expand_px: Optional[Numeric]
+):
     """Spot detection by intensity filter
 
     Arguments
@@ -61,16 +71,16 @@ def detect_spots_int(*, input_image, spot_threshold: Numeric, expand_px: Optiona
     spot_threshold : NumberLike
         Minimum intensity value in a pixel to consider it as part of a spot region
     expand_px : float or int or NoneType
-        Number of pixels by which to expand contiguous subregion, 
+        Number of pixels by which to expand contiguous subregion,
         up to point of overlap with neighboring subregion of image
-        
+
     Returns
     -------
-    pd.DataFrame, np.ndarray, np.ndarray: 
-        The centroids and roi_IDs of the spots found, 
-        the image used for spot detection, and 
-        numpy array with only sufficiently large regions 
-        retained (bigger than threshold number of pixels), 
+    pd.DataFrame, np.ndarray, np.ndarray:
+        The centroids and roi_IDs of the spots found,
+        the image used for spot detection, and
+        numpy array with only sufficiently large regions
+        retained (bigger than threshold number of pixels),
         and dilated by expansion amount (possibly)
     """
     # TODO: enforce that output column names don't vary with code path walked.
@@ -82,34 +92,51 @@ def detect_spots_int(*, input_image, spot_threshold: Numeric, expand_px: Optiona
     labels, num_obj = ndi.label(binary, structure=struct)
     if num_obj > 1:
         labels = remove_small_objects(labels, min_size=5)
-    spot_props, labels = _build_props_table(labels=labels, input_image=input_image, expand_px=expand_px)
+    spot_props, labels = _build_props_table(
+        labels=labels, input_image=input_image, expand_px=expand_px
+    )
     return spot_props, input_image, labels
 
 
-def _build_props_table(*, labels: np.ndarray, input_image: np.ndarray, expand_px: Optional[int]) -> Tuple[pd.DataFrame, np.ndarray]:
+def _build_props_table(
+    *, labels: np.ndarray, input_image: np.ndarray, expand_px: Optional[int]
+) -> Tuple[pd.DataFrame, np.ndarray]:
     if expand_px:
         labels = expand_labels(labels, expand_px)
     if np.all(labels == 0):
         # No substructures (ROIs) exist.
-        spot_props = pd.DataFrame(columns=['label', 'centroid_weighted-0', 'centroid_weighted-1', 'centroid_weighted-2', 'area', 'intensity_mean'])
+        spot_props = pd.DataFrame(
+            columns=[
+                "label",
+                "centroid_weighted-0",
+                "centroid_weighted-1",
+                "centroid_weighted-2",
+                "area",
+                "intensity_mean",
+            ]
+        )
     else:
         spot_props = regionprops_table(
-            label_image=labels, 
-            intensity_image=input_image, 
-            properties=('label', 'centroid_weighted', 'area', 'intensity_mean'),
-            )
+            label_image=labels,
+            intensity_image=input_image,
+            properties=("label", "centroid_weighted", "area", "intensity_mean"),
+        )
         spot_props = pd.DataFrame(spot_props)
-    spot_props = spot_props.drop(['label'], axis=1)
+    spot_props = spot_props.drop(["label"], axis=1)
     spot_props = spot_props.rename(columns=CENTROID_COLUMNS_REMAPPING)
     spot_props = spot_props.reset_index(drop=True)
     return spot_props, labels
 
 
 def _check_input_image(img: np.ndarray) -> NoReturn:
     if not isinstance(img, np.ndarray):
-        raise TypeError(f"Expected numpy array for input image but got {type(img).__name__}")
+        raise TypeError(
+            f"Expected numpy array for input image but got {type(img).__name__}"
+        )
     if img.ndim != 3:
-        raise DimensionalityError(f"Expected 3D input image but got {img.ndim}-dimensional")
+        raise DimensionalityError(
+            f"Expected 3D input image but got {img.ndim}-dimensional"
+        )
 
 
 def _preprocess_for_difference_of_gaussians(input_image: np.ndarray) -> np.ndarray:

diff --git a/tests/test_old_new_equivalence.py b/tests/test_old_new_equivalence.py
@@ -1,10 +1,11 @@
 import os
-import numpy as np
 from pathlib import Path
+
+import numpy as np
 import pytest
 
-from spotfishing.deprecated import detect_spots_int_old, detect_spots_dog_old
-from spotfishing import detect_spots_int, detect_spots_dog
+from spotfishing import detect_spots_dog, detect_spots_int
+from spotfishing.deprecated import detect_spots_dog_old, detect_spots_int_old
 
 OLD_PIXEL_EXPANSION_INT = 1
 OLD_PIXEL_EXPANSION_DOG = 10
@@ -14,18 +15,32 @@ def get_img_data_file(fn: str) -> Path:
     return Path(os.path.dirname(__file__)) / "data" / fn
 
 
-@pytest.mark.parametrize("data_path", [get_img_data_file(fn) for fn in ("img__p0_t57_c0__smaller.npy", "img__p13_t57_c0__smaller.npy")])
 @pytest.mark.parametrize(
-    ["old_fun", "new_fun", "threshold", "expand_px"], 
-    [(detect_spots_int_old, detect_spots_int, threshold, OLD_PIXEL_EXPANSION_INT) for threshold in (500, 300, 100)] + 
-    [(detect_spots_dog_old, detect_spots_dog, threshold , OLD_PIXEL_EXPANSION_DOG) for threshold in (20, 15, 10)]
+    "data_path",
+    [
+        get_img_data_file(fn)
+        for fn in ("img__p0_t57_c0__smaller.npy", "img__p13_t57_c0__smaller.npy")
+    ],
+)
+@pytest.mark.parametrize(
+    ["old_fun", "new_fun", "threshold", "expand_px"],
+    [
+        (detect_spots_int_old, detect_spots_int, threshold, OLD_PIXEL_EXPANSION_INT)
+        for threshold in (500, 300, 100)
+    ]
+    + [
+        (detect_spots_dog_old, detect_spots_dog, threshold, OLD_PIXEL_EXPANSION_DOG)
+        for threshold in (20, 15, 10)
+    ],
 )
 def test_eqv(data_path, old_fun, new_fun, threshold, expand_px):
     data = np.load(data_path)
     old_table, _, _ = old_fun(data, threshold)
-    new_table, _, _ = new_fun(input_image=data, spot_threshold=threshold, expand_px=expand_px)
+    new_table, _, _ = new_fun(
+        input_image=data, spot_threshold=threshold, expand_px=expand_px
+    )
     assert np.all(old_table.index == new_table.index)
     cols = ["zc", "yc", "xc", "area", "intensity_mean"]
-    sub_cols = [c for c in cols if c != "area"] # wasn't in DoG before
+    sub_cols = [c for c in cols if c != "area"]  # wasn't in DoG before
     assert list(new_table.columns) == cols
     assert np.all(old_table[sub_cols].to_numpy() == new_table[sub_cols].to_numpy())