Integration tests

.github/workflows/build.yml

@@ -39,8 +39,29 @@ jobs:
           path: ${{github.workspace}}/dist/raidionicsseg-*.whl
           if-no-files-found: error
 
+  setup-test-data:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: 3.9
+
+      - name: Download test resources
+        working-directory: tests
+        run: |
+            pip install requests
+            python -c "from download_resources import download_resources; download_resources('../test_data')"
+
+      - name: Upload test resources
+        uses: actions/upload-artifact@v4
+        with:
+          name: test-resources
+          path: ./test_data
   test:
-    needs: build
+    needs: [build, setup-test-data]
     runs-on: ${{ matrix.os }}
     strategy:
       matrix:
@@ -127,6 +148,12 @@ jobs:
         - name: Clone repo
           uses: actions/checkout@v4
 
+        - name: Download test resources
+          uses: actions/download-artifact@v4
+          with:
+            name: test-resources
+            path: ./tests/unit_tests_results_dir
+
         - name: Integration tests
           run: |
               pip install pytest pytest-cov pytest-timeout requests

.github/workflows/build_macos_arm_11.yml

@@ -76,7 +76,7 @@ jobs:
           run: cd ${{github.workspace}}/tests && python3 test_inference_segmentation_mediastinum.py
 
         - name: Inference unit test with test-time augmentation
-          run: cd ${{github.workspace}}/tests && python3 test_inference_segmentation_test_time_augmentation.py
+          run: cd ${{github.workspace}}/tests && python3 test_inference_segmentation_advanced.py
 
 #        - name: Test with pytest
 #          run: |

pyproject.toml

@@ -15,7 +15,7 @@ dependencies = [
   "nibabel",
   "h5py",
   "pandas",
-  "SimpleITK",
+  "SimpleITK<=2.4.1",
   "aenum",
   "scikit-image",
   "tqdm",

raidionicsseg/PreProcessing/brain_clipping.py

@@ -1,11 +1,10 @@
-import configparser
 import logging
 import os
-import subprocess
 from copy import deepcopy
 from pathlib import PurePath
 from typing import List
 from typing import Tuple
+from typing import Union
 
 import numpy as np
 from nibabel.processing import resample_to_output
@@ -18,12 +17,10 @@
 
 
 def crop_MR_background(
-    filepath: str,
     volume: np.ndarray,
     new_spacing: Tuple[float],
-    storage_path: str,
     parameters: ConfigResources,
-    crop_bbox=None,
+    crop_bbox: List[int] = None,
 ) -> Tuple[np.ndarray, List[int]]:
     """
     Performs different background cropping inside an MRI volume, as defined by the 'crop_background' stored in a model
@@ -53,12 +50,14 @@ def crop_MR_background(
         The bounding region is expressed as: [minx, miny, minz, maxx, maxy, maxz].
     """
     if parameters.crop_background == "minimum":
-        return crop_MR(volume, parameters, crop_bbox)
+        return crop_background_minimum(volume=volume, crop_bbox=crop_bbox)
     elif parameters.crop_background == "brain_clip" or parameters.crop_background == "brain_mask":
-        return skull_stripping_tf(filepath, volume, new_spacing, storage_path, parameters)
+        return skull_stripping(volume=volume, new_spacing=new_spacing, parameters=parameters)
 
 
-def crop_MR(volume: np.ndarray, parameters, crop_bbox=None) -> Tuple[np.ndarray, List[int]]:
+def crop_background_minimum(
+    volume: np.ndarray, crop_bbox: Union[None, List[int]] = None
+) -> Tuple[np.ndarray, List[int]]:
     """
     Performs background cropping inside an MRI volume in 'minimum' mode whereby the black space around
     the head is removed.
@@ -67,8 +66,7 @@ def crop_MR(volume: np.ndarray, parameters, crop_bbox=None) -> Tuple[np.ndarray,
     ----------
     volume : np.ndarray
         Patient MRI volume to crop.
-    parameters : :obj:`ConfigResources`
-        UNUSED, to remove.
+    crop_bbox:
     Returns
     -------
     np.ndarray
@@ -107,20 +105,18 @@ def crop_MR(volume: np.ndarray, parameters, crop_bbox=None) -> Tuple[np.ndarray,
     return cropped_volume, crop_bbox
 
 
-def skull_stripping_tf(
-    filepath, volume: np.ndarray, new_spacing: Tuple[float], storage_path: str, parameters: ConfigResources
+def skull_stripping(
+    volume: np.ndarray, new_spacing: Tuple[float], parameters: ConfigResources
 ) -> Tuple[np.ndarray, List[int]]:
     """
-    Generates a brain segmentation mask by running model inference and performs skull stripping.
+    Performs skull stripping over the provided input volume using the brain mask manually provided.
 
     Parameters
     ----------
     volume : np.ndarray
         Patient MRI volume to skull strip.
     new_spacing : Tuple[float]
         .
-    storage_path : str
-        Destination folder where the results will be stored.
     parameters :  :obj:`ConfigResources`
         Loaded configuration specifying runtime parameters.
     Returns
@@ -132,23 +128,7 @@ def skull_stripping_tf(
         The bounding region is expressed as: [minx, miny, minz, maxx, maxy, maxz].
     """
     if not os.path.exists(parameters.runtime_brain_mask_filepath):
-        brain_config_filename = os.path.join(os.path.dirname(parameters.config_filename), "brain_main_config.ini")
-        new_parameters = configparser.ConfigParser()
-        new_parameters.read(parameters.config_filename)
-        new_parameters.set(
-            "System", "model_folder", os.path.join(os.path.dirname(parameters.model_folder), "MRI_Brain")
-        )
-        new_parameters.set("Runtime", "reconstruction_method", "thresholding")
-        new_parameters.set("Runtime", "reconstruction_order", "resample_first")
-        with open(brain_config_filename, "w") as cf:
-            new_parameters.write(cf)
-        old_parameters = deepcopy(parameters)
-        from raidionicsseg.fit import run_model
-
-        run_model(brain_config_filename)
-        brain_mask_filename = os.path.join(storage_path, "labels_Brain.nii.gz")
-        os.remove(brain_config_filename)
-        parameters = old_parameters
+        raise ValueError("A brain segmentation mask must be provided inside ['Neuro']['brain_segmentation_filename']")
     else:
         brain_mask_filename = parameters.runtime_brain_mask_filepath
 
@@ -176,7 +156,7 @@ def advanced_crop_exclude_background(
     volume: np.ndarray, crop_mode: str, brain_mask: np.ndarray
 ) -> Tuple[np.ndarray, List[int]]:
     """
-    Perfoms skull stripping either in mode 'brain_clip' or mode 'brain_mask'.
+    Performs skull stripping either in mode 'brain_clip' or mode 'brain_mask'.
 
     Parameters
     ----------

raidionicsseg/PreProcessing/mediastinum_clipping.py

@@ -1,4 +1,4 @@
-import configparser
+import logging
 import os
 from copy import deepcopy
 from typing import List
@@ -15,7 +15,7 @@
 
 
 def crop_mediastinum_volume(
-    filepath: str, volume: np.ndarray, new_spacing: Tuple[float], storage_path: str, parameters: ConfigResources
+    volume: np.ndarray, new_spacing: Tuple[float], parameters: ConfigResources, crop_bbox: List[int] = None
 ) -> Tuple[np.ndarray, List[int]]:
     """
     Performs different background cropping inside a mediastinal CT volume, as defined by the 'crop_background' stored
@@ -25,14 +25,10 @@ def crop_mediastinum_volume(
 
     Parameters
     ----------
-    filepath : str
-        Filepath of the input volume (CT or MRI) to use.
     volume : np.ndarray
         .
     new_spacing : Tuple[float]
         .
-    storage_path : str
-        Destination folder where the results will be stored.
     parameters :  :obj:`ConfigResources`
         Loaded configuration specifying runtime parameters.
     Returns
@@ -45,90 +41,110 @@ def crop_mediastinum_volume(
         The bounding region is expressed as: [minx, miny, minz, maxx, maxy, maxz].
     """
     if parameters.crop_background == "minimum":
-        return mediastinum_clipping(volume, parameters)
-    elif parameters.crop_background == "brain_clip" or parameters.crop_background == "brain_mask":
-        return mediastinum_clipping_DL(filepath, volume, new_spacing, storage_path, parameters)
-
-
-def mediastinum_clipping(volume, parameters):
-    intensity_threshold = -250
-    airmetal_mask = deepcopy(volume)
-    airmetal_mask[airmetal_mask > intensity_threshold] = 0
-    airmetal_mask[airmetal_mask <= intensity_threshold] = 1
-
-    airmetal_mask = smo.binary_closing(airmetal_mask, iterations=5)
-
-    labels, nb_components = smeas.label(airmetal_mask)
-    airmetal_pieces = smeas.find_objects(labels, min(nb_components, 1000))
-
-    nums = []
-    for p in enumerate(airmetal_pieces):
-        bb = p[1]
-        z = bb[2].stop - bb[2].start
-        y = bb[1].stop - bb[1].start
-        x = bb[0].stop - bb[0].start
-        nums.append(x * y * z)
-
-    # Should check if the first two or three elements are "as big". If two the following code is correct, if three
-    # something should be changed so that the lungs is the third (normally?) and background the first two.
-    ind_bg = nums.index(np.max(nums)) + 1
-    nums.remove(np.max(nums))
-    ind_lungs = (
-        nums.index(np.max(nums)) + 1 + 1
-    )  # +1 because find_objects labels start at 1, and +1 because we remove one value above
-    # ind_lungs = nums.index(sorted(nums, reverse=True)[0])+1
-    # for l in range(nb_components):
-    #   nums.append(np.count_nonzero(np.where(labels == l)))
-
-    background_mask = np.copy(labels)
-    background_mask[background_mask != ind_bg] = 0
-
-    lungstrachea_mask = np.copy(labels)
-    lungstrachea_mask[lungstrachea_mask != ind_lungs] = 0
-    lungstrachea_mask[lungstrachea_mask == ind_lungs] = 1
-
-    lungs_boundingbox = airmetal_pieces[ind_lungs - 1]  # Because indexing starts at 0, so have to decrease by one
-    crop_bbox = [
-        lungs_boundingbox[0].start,
-        lungs_boundingbox[1].start,
-        lungs_boundingbox[2].start,
-        lungs_boundingbox[0].stop,
-        lungs_boundingbox[1].stop,
-        lungs_boundingbox[2].stop,
-    ]
-
-    cropped_volume = volume[crop_bbox[0] : crop_bbox[3], crop_bbox[1] : crop_bbox[4], crop_bbox[2] : crop_bbox[5]]
-
-    print("Cropped mediastinum values: {}".format(lungs_boundingbox))
+        return mediastinum_clipping(volume=volume, parameters=parameters, crop_bbox=crop_bbox)
+    elif parameters.crop_background == "lungs_clip" or parameters.crop_background == "lungs_mask":
+        return mediastinum_clipping_advanced(
+            volume=volume, new_spacing=new_spacing, parameters=parameters
+        )
+
+
+def mediastinum_clipping(
+    volume: np.ndarray, parameters: ConfigResources, crop_bbox=None
+) -> Tuple[np.ndarray, List[int]]:
+    if crop_bbox is None:
+        intensity_threshold = -250
+        airmetal_mask = deepcopy(volume)
+        airmetal_mask[airmetal_mask > intensity_threshold] = 0
+        airmetal_mask[airmetal_mask <= intensity_threshold] = 1
+
+        airmetal_mask = smo.binary_closing(airmetal_mask, iterations=5)
+
+        labels, nb_components = smeas.label(airmetal_mask)
+        airmetal_pieces = smeas.find_objects(labels, min(nb_components, 1000))
+
+        nums = []
+        for p in enumerate(airmetal_pieces):
+            bb = p[1]
+            z = bb[2].stop - bb[2].start
+            y = bb[1].stop - bb[1].start
+            x = bb[0].stop - bb[0].start
+            nums.append(x * y * z)
+
+        # Should check if the first two or three elements are "as big". If two the following code is correct, if three
+        # something should be changed so that the lungs is the third (normally?) and background the first two.
+        ind_bg = nums.index(np.max(nums)) + 1
+        nums.remove(np.max(nums))
+        ind_lungs = (
+            nums.index(np.max(nums)) + 1 + 1
+        )  # +1 because find_objects labels start at 1, and +1 because we remove one value above
+        # ind_lungs = nums.index(sorted(nums, reverse=True)[0])+1
+        # for l in range(nb_components):
+        #   nums.append(np.count_nonzero(np.where(labels == l)))
+
+        background_mask = np.copy(labels)
+        background_mask[background_mask != ind_bg] = 0
+
+        lungstrachea_mask = np.copy(labels)
+        lungstrachea_mask[lungstrachea_mask != ind_lungs] = 0
+        lungstrachea_mask[lungstrachea_mask == ind_lungs] = 1
+
+        lungs_boundingbox = airmetal_pieces[ind_lungs - 1]  # Because indexing starts at 0, so have to decrease by one
+        crop_bbox = [
+            lungs_boundingbox[0].start,
+            lungs_boundingbox[1].start,
+            lungs_boundingbox[2].start,
+            lungs_boundingbox[0].stop,
+            lungs_boundingbox[1].stop,
+            lungs_boundingbox[2].stop,
+        ]
+
+        cropped_volume = volume[crop_bbox[0] : crop_bbox[3], crop_bbox[1] : crop_bbox[4], crop_bbox[2] : crop_bbox[5]]
+    else:
+        min_row, min_col, min_depth, max_row, max_col, max_depth = (
+            crop_bbox[0],
+            crop_bbox[1],
+            crop_bbox[2],
+            crop_bbox[3],
+            crop_bbox[4],
+            crop_bbox[5],
+        )
+        cropped_volume = volume[min_row:max_row, min_col:max_col, min_depth:max_depth]
+    logging.debug(
+        "Mediastinum background cropping with: [{}, {}, {}, {}, {}, {}].\n".format(
+            crop_bbox[0], crop_bbox[1], crop_bbox[2], crop_bbox[3], crop_bbox[4], crop_bbox[5]
+        )
+    )
     return cropped_volume, crop_bbox
 
 
-def mediastinum_clipping_DL(filepath, volume, new_spacing, storage_path, parameters):
+def mediastinum_clipping_advanced(
+    volume: np.ndarray, new_spacing: Tuple[float], parameters: ConfigResources
+) -> Tuple[np.ndarray, List[int]]:
+    """
+
+    Parameters
+    ----------
+    volume
+    new_spacing
+    parameters
+
+    Returns
+    -------
+
+    """
     if not parameters.runtime_lungs_mask_filepath and not os.path.exists(parameters.runtime_lungs_mask_filepath):
-        lung_config_filename = os.path.join(os.path.dirname(parameters.config_filename), "lungs_main_config.ini")
-        new_parameters = configparser.ConfigParser()
-        new_parameters.read(parameters.config_filename)
-        new_parameters.set("System", "model_folder", os.path.join(os.path.dirname(parameters.model_folder), "CT_Lungs"))
-        new_parameters.set("Runtime", "reconstruction_method", "thresholding")
-        new_parameters.set("Runtime", "reconstruction_order", "resample_first")
-        with open(lung_config_filename, "w") as cf:
-            new_parameters.write(cf)
-        old_parameters = deepcopy(parameters)
-        from raidionicsseg.fit import run_model
-
-        run_model(lung_config_filename)
-        lungs_mask_filename = os.path.join(storage_path, "labels_Lungs.nii.gz")
-        os.remove(lung_config_filename)
-        parameters = old_parameters
+        raise ValueError(
+            "A brain segmentation mask must be provided inside ['Mediastinum']['lungs_segmentation_filename']"
+        )
     else:
         lungs_mask_filename = parameters.runtime_lungs_mask_filepath
 
     lungs_mask_ni = load_nifti_volume(lungs_mask_filename)
     resampled_volume = resample_to_output(lungs_mask_ni, new_spacing, order=0)
     lungs_mask = resampled_volume.get_fdata().astype("uint8")
+
     # In case the lungs mask has a different label for each lung
     lungs_mask[lungs_mask != 0] = 1
-
     lung_region = regionprops(lungs_mask)
     min_row, min_col, min_depth, max_row, max_col, max_depth = lung_region[0].bbox
     if parameters.crop_background == "invert":