Merge branch 'master' into new_som

README.md

@@ -99,3 +99,8 @@ If you run into trouble, please first refer to our [FAQ](https://ark-analysis.re
 ## Want to contribute?  
 
 If you would like to help make `ark` better, please take a look at our [contributing guidelines](https://ark-analysis.readthedocs.io/en/latest/_rtd/contributing.html). 
+
+## Citation
+Please cite our paper if you found our repo useful!
+
+[Greenwald, Miller et al. Whole-cell segmentation of tissue images with human-level performance using large-scale data annotation and deep learning](https://www.biorxiv.org/content/10.1101/2021.03.01.431313v2)

ark/segmentation/marker_quantification.py

@@ -255,7 +255,7 @@ def compute_marker_counts(input_images, segmentation_labels, nuclear_counts=Fals
                                                       nuc_segmentation_labels=nuc_labels,
                                                       cell_ids=unique_cell_ids)
 
-        nuc_props = get_single_compartment_props(segmentation_labels.loc[:, :, 'nuclear'].values,
+        nuc_props = get_single_compartment_props(nuc_labels,
                                                  regionprops_base, regionprops_single_comp,
                                                  **reg_props)
 

ark/segmentation/marker_quantification_test.py

@@ -315,6 +315,25 @@ def test_compute_marker_counts_nuc_whole_cell_diff():
     assert np.array_equal(segmentation_output_unequal.loc['nuclear', :, 'cell_size'],
                           segmentation_output_unequal.loc['nuclear', :, 'area'])
 
+    # check that splitting large nuclei works as expected
+
+    # swap nuclear and cell masks so that nuc is bigger
+    big_nuc_masks = np.concatenate((nuc_mask, cell_mask), axis=-1)
+    segmentation_labels_big_nuc = test_utils.make_labels_xarray(
+        label_data=big_nuc_masks,
+        compartment_names=['whole_cell', 'nuclear']
+    )
+
+    # test utils output is 4D but tests require 3D
+    segmentation_labels_big_nuc = segmentation_labels_big_nuc[0]
+
+    segmentation_output_big_nuc = \
+        marker_quantification.compute_marker_counts(
+            input_images=input_images,
+            segmentation_labels=segmentation_labels_big_nuc,
+            nuclear_counts=True,
+            split_large_nuclei=True)
+
 
 def test_compute_marker_counts_no_coords():
     cell_mask, channel_data = test_utils.create_test_extraction_data()

ark/utils/io_utils.py

@@ -3,12 +3,14 @@
 import warnings
 
 
-def validate_paths(paths):
+def validate_paths(paths, data_prefix=True):
     """Verifys that paths exist and don't leave Docker's scope
 
     Args:
         paths (str or list):
             paths to verify.
+        data_prefix (bool):
+            if True, checks that directory starts with /data, necessary when inside the docker
 
     Raises:
         ValueError:
@@ -21,7 +23,7 @@ def validate_paths(paths):
 
     for path in paths:
         if not os.path.exists(path):
-            if str(path).startswith('../data'):
+            if str(path).startswith('../data') or not data_prefix:
                 for parent in reversed(pathlib.Path(path).parents):
                     if not os.path.exists(parent):
                         raise ValueError(

ark/utils/io_utils_test.py

@@ -52,6 +52,23 @@ def test_validate_paths():
         with pytest.raises(ValueError, match=r".*The file/path.*not_a_real_file.*"):
             iou.validate_paths(wrong_file)
 
+    # make tempdir for testing outside of docker
+    with tempfile.TemporaryDirectory() as valid_path:
+
+        # make valid subdirectory
+        valid_parts = [p for p in pathlib.Path(valid_path).parts]
+        valid_parts[0] = 'not_a_real_directory'
+
+        # test no '../data' prefix
+        starts_out_of_scope = os.path.join(*valid_parts)
+
+        # test out of scope when specifying out of scope
+        with pytest.raises(ValueError, match=r".*not_a_real_directory*"):
+            iou.validate_paths(starts_out_of_scope, data_prefix=False)
+
+        with pytest.raises(ValueError, match=r".*is not prefixed with.*"):
+            iou.validate_paths(starts_out_of_scope, data_prefix=True)
+
     # reset cwd after testing
     os.chdir('../')
 

ark/utils/load_utils.py

@@ -35,7 +35,7 @@ def load_imgs_from_mibitiff(data_dir, mibitiff_files=None, channels=None, delimi
             xarray with shape [fovs, x_dim, y_dim, channels]
     """
 
-    iou.validate_paths(data_dir)
+    iou.validate_paths(data_dir, data_prefix=False)
 
     if not mibitiff_files:
         mibitiff_files = iou.list_files(data_dir, substrs=['.tif'])
@@ -108,7 +108,7 @@ def load_imgs_from_tree(data_dir, img_sub_folder=None, fovs=None, channels=None,
             xarray with shape [fovs, x_dim, y_dim, tifs]
     """
 
-    iou.validate_paths(data_dir)
+    iou.validate_paths(data_dir, data_prefix=False)
 
     if fovs is None:
         # get all fovs
@@ -237,7 +237,7 @@ def load_imgs_from_dir(data_dir, files=None, delimiter=None, xr_dim_name='compar
               of channels in the input.
     """
 
-    iou.validate_paths(data_dir)
+    iou.validate_paths(data_dir, data_prefix=False)
 
     if files is None:
         imgs = iou.list_files(data_dir, substrs=['.tif', '.jpg', '.png'])

ark/utils/segmentation_utils.py

@@ -39,7 +39,7 @@ def find_nuclear_label_id(nuc_segmentation_labels, cell_coords):
     return nuclear_label_id
 
 
-def split_large_nuclei(cell_segmentation_labels, nuc_segmentation_labels, cell_ids, min_size=5):
+def split_large_nuclei(cell_segmentation_labels, nuc_segmentation_labels, cell_ids, min_size=15):
     """Splits nuclei that are bigger than the corresponding cell into multiple pieces
 
     Args:

ark/utils/spatial_analysis_utils.py

@@ -31,7 +31,7 @@ def calc_dist_matrix(label_maps, save_path=None):
     # Check that file path exists, if given
 
     if save_path is not None:
-        io_utils.validate_paths(save_path)
+        io_utils.validate_paths(save_path, data_prefix=False)
 
     dist_mats_list = []