Background MIPs for TFE

nuc_morph_analysis/lib/visualization/write_data_for_colorizer.py

@@ -22,6 +22,10 @@
 from nuc_morph_analysis.lib.preprocessing.global_dataset_filtering import (
     load_dataset_with_features,
 )
+from nuc_morph_analysis.lib.visualization.write_mips_for_colorizer import (
+    save_colony_backdrop_mips
+)
+
 from nuc_morph_analysis.lib.visualization.plotting_tools import (
     get_plot_labels_for_metric,
 )
@@ -68,13 +72,7 @@ class NucMorphFeatureSpec:
 # volume	float	In FMS manifest	Volume of a single nucleus in pixels in a given frame
 # height	float	In FMS manifest	Height (in the z-direction) of the a single nucleus in pixels in a given frame
 # NUC_PC1	float	Needs calculated and added	Value for shape mode 1 for a single nucleus in a given frame
-# NUC_PC2	float	Needs calculated and added	Value for shape mode 2 for a single nucleus in a given frame
-# NUC_PC3	float	Needs calculated and added	Value for shape mode 3 for a single nucleus in a given frame
-# NUC_PC4	float	Needs calculated and added	Value for shape mode 4 for a single nucleus in a given frame
-# NUC_PC5	float	Needs calculated and added	Value for shape mode 5 for a single nucleus in a given frame
-# NUC_PC6	float	Needs calculated and added	Value for shape mode 6 for a single nucleus in a given frame
-# NUC_PC7	float	Needs calculated and added	Value for shape mode 7 for a single nucleus in a given frame
-# NUC_PC8	float	Needs calculated and added	Value for shape mode 8 for a single nucleus in a given frame
+
 
 
 OBJECT_ID_COLUMN = "label_img"
@@ -231,26 +229,26 @@ class NucMorphFeatureSpec:
         NucMorphFeatureSpec("SA_vol_ratio"),
 
         # mitotic and apoptotic neighbor columns
-        NucMorphFeatureSpec(column_name="frame_of_breakdown", type=FeatureType.CATEGORICAL, categories=["False", "True"]),
-        NucMorphFeatureSpec(column_name="frame_of_formation", type=FeatureType.CATEGORICAL, categories=["False", "True"]),
-        NucMorphFeatureSpec(column_name="number_of_frame_of_breakdown_neighbors"),
-        NucMorphFeatureSpec(column_name="number_of_frame_of_formation_neighbors"),
-        NucMorphFeatureSpec(column_name="has_mitotic_neighbor_breakdown", type=FeatureType.CATEGORICAL, categories=["False", "True"]),
-        NucMorphFeatureSpec(column_name="has_mitotic_neighbor_formation", type=FeatureType.CATEGORICAL, categories=["False", "True"]),
-        NucMorphFeatureSpec(column_name="has_mitotic_neighbor_breakdown_forward_dilated", type=FeatureType.CATEGORICAL, categories=["False", "True"]),
-        NucMorphFeatureSpec(column_name="has_mitotic_neighbor_formation_backward_dilated", type=FeatureType.CATEGORICAL, categories=["False", "True"]),
-        NucMorphFeatureSpec(column_name="has_mitotic_neighbor", type=FeatureType.CATEGORICAL, categories=["False", "True"]),
-        NucMorphFeatureSpec(column_name="has_mitotic_neighbor_dilated", type=FeatureType.CATEGORICAL, categories=["False", "True"]),
-        NucMorphFeatureSpec(column_name="frame_of_death", type=FeatureType.CATEGORICAL, categories=["False", "True"]),
-        NucMorphFeatureSpec(column_name="has_dying_neighbor", type=FeatureType.CATEGORICAL, categories=["False", "True"]),
-        NucMorphFeatureSpec(column_name="has_dying_neighbor_forward_dilated", type=FeatureType.CATEGORICAL, categories=["False", "True"]),
-        NucMorphFeatureSpec(column_name="number_of_frame_of_death_neighbors"),
-        NucMorphFeatureSpec(column_name="sum_has_mitotic_neighbor_breakdown"), # per track feature
-        NucMorphFeatureSpec(column_name="sum_has_mitotic_neighbor_formation"),# per track feature
-        NucMorphFeatureSpec(column_name="sum_has_mitotic_neighbor"),# per track feature
-        NucMorphFeatureSpec(column_name="sum_has_dying_neighbor"),# per track feature
-        NucMorphFeatureSpec(column_name="sum_number_of_frame_of_breakdown_neighbors"),# per track feature
-        NucMorphFeatureSpec(column_name="number_of_frame_of_death_neighbors"),# per track feature
+        NucMorphFeatureSpec("frame_of_breakdown", type=FeatureType.CATEGORICAL, categories=["False", "True"]),
+        NucMorphFeatureSpec("frame_of_formation", type=FeatureType.CATEGORICAL, categories=["False", "True"]),
+        NucMorphFeatureSpec("number_of_frame_of_breakdown_neighbors"),
+        NucMorphFeatureSpec("number_of_frame_of_formation_neighbors"),
+        NucMorphFeatureSpec("has_mitotic_neighbor_breakdown", type=FeatureType.CATEGORICAL, categories=["False", "True"]),
+        NucMorphFeatureSpec("has_mitotic_neighbor_formation", type=FeatureType.CATEGORICAL, categories=["False", "True"]),
+        NucMorphFeatureSpec("has_mitotic_neighbor_breakdown_forward_dilated", type=FeatureType.CATEGORICAL, categories=["False", "True"]),
+        NucMorphFeatureSpec("has_mitotic_neighbor_formation_backward_dilated", type=FeatureType.CATEGORICAL, categories=["False", "True"]),
+        NucMorphFeatureSpec("has_mitotic_neighbor", type=FeatureType.CATEGORICAL, categories=["False", "True"]),
+        NucMorphFeatureSpec("has_mitotic_neighbor_dilated", type=FeatureType.CATEGORICAL, categories=["False", "True"]),
+        NucMorphFeatureSpec("frame_of_death", type=FeatureType.CATEGORICAL, categories=["False", "True"]),
+        NucMorphFeatureSpec("has_dying_neighbor", type=FeatureType.CATEGORICAL, categories=["False", "True"]),
+        NucMorphFeatureSpec("has_dying_neighbor_forward_dilated", type=FeatureType.CATEGORICAL, categories=["False", "True"]),
+        NucMorphFeatureSpec("number_of_frame_of_death_neighbors"),
+        NucMorphFeatureSpec("sum_has_mitotic_neighbor_breakdown"), # per track feature
+        NucMorphFeatureSpec("sum_has_mitotic_neighbor_formation"),# per track feature
+        NucMorphFeatureSpec("sum_has_mitotic_neighbor"),# per track feature
+        NucMorphFeatureSpec("sum_has_dying_neighbor"),# per track feature
+        NucMorphFeatureSpec("sum_number_of_frame_of_breakdown_neighbors"),# per track feature
+        NucMorphFeatureSpec("number_of_frame_of_death_neighbors"),# per track feature
 
 
         # new columns
@@ -419,7 +417,7 @@ def make_features(
         if scale_factor is not None:
             data = data * scale_factor
 
-        description = GLOSSARY[feature.column_name]
+        description = GLOSSARY.get(feature.column_name, "")
 
         writer.write_feature(
             data,
@@ -447,6 +445,7 @@ def make_dataset(
     do_frames=True,
     scale=0.25,
     parallel=False,
+    make_backdrops=True,
 ):
     """Make a new dataset from the given data, and write the complete dataset
     files to the given output directory.
@@ -459,11 +458,11 @@ def make_dataset(
 
     # load the dataset once
     df_all = load_dataset_with_features("all_baseline", remove_growth_outliers=False)
-
+    
     for filter in filters:
         output_dir_subset = Path(output_dir) / filter
         output_dir_subset.mkdir(parents=True, exist_ok=True)
-        output_dir_subset = str(output_dir_subset)
+        output_dir_subset = str(output_dir_subset)        
 
         df_filter = df_all.copy()
         df_filter.loc[df_filter[filter] == False, "is_outlier"] = True
@@ -511,12 +510,19 @@ def make_dataset(
             )
 
             # Make the features, frame data, and manifest.
-            nframes = len(grouped_frames)
+            nframes = len(grouped_frames)         
             writer.set_frame_paths(generate_frame_paths(nframes))
 
             make_features(full_dataset, FEATURE_COLUMNS[filter], dataset, writer)
+
             if do_frames:
                 make_all_frames(grouped_frames, scale, writer, parallel)
+
+            if make_backdrops:
+                save_colony_backdrop_mips(dataset, output_dir_subset + f"/{dataset}/backdrops/")     
+                backdrop_paths = [f"./backdrops/{i}.png" for i in range(nframes)]
+                writer.add_backdrops("Max intensity z-projection of Lamin B1", backdrop_paths)
+
             writer.write_manifest(metadata=metadata)
 
 
@@ -558,6 +564,12 @@ def make_dataset(
     "Has no effect with --noframes.",
 )
 
+parser.add_argument(
+    "--make_backdrops",
+    type=bool,
+    default=True,
+    help="True will generate backdrops (Default). False will use previously generated backdrops or display none if they have not been generated before.",
+)
 args = parser.parse_args()
 
 
@@ -572,6 +584,7 @@ def main():
         do_frames=not args.noframes,
         scale=args.scale,
         parallel=args.parallel,
+        make_backdrops=args.make_backdrops,
     )
 
 

nuc_morph_analysis/lib/visualization/write_mips_for_colorizer.py

@@ -0,0 +1,58 @@
+#%%
+import os
+import numpy as np
+import skimage.exposure as skex
+from skimage.transform import resize
+from aicsimageio.writers import two_d_writer
+from nuc_morph_analysis.lib.preprocessing.single_track_contact.export_code import export_helper
+from nuc_morph_analysis.lib.preprocessing import load_data
+
+# %%
+def save_colony_backdrop_mips(colony, figdir, dtype ="uint8", downsample_factor=2, zarr_res_level=0):
+    """
+    Save maximum intensity projection (MIP) images for a given colony.
+
+    This function reads the raw field of view images for the specified colony. The images are
+    processed to flip them to match the orientation of the segmentation images on TFE, intensity 
+    rescaled, downsampled and max intensity projections in z are saved as png files to a directory.
+    
+    Two downsampling methods are available. 
+    
+    1. Load directly a precomputed lower resolution zarr. Options are 0, 1, 2, 3, 4.
+    2. Resize the image using skimage and downsample the image using a custom scaling factor. 
+    
+    The default behavior is to resize the image using skimage. This is the same method used to generate
+    MIPs in Figure 1. 
+
+    Parameters:
+    -----------
+    colony (str): The name of the colony to process.
+    figdir (str): The directory where the MIP images will be saved.
+    dtype (str): The data type to use for the images. Default is "uint8".
+    downsample_factor (int): The factor by which to downsample the images. Default is 2. 
+    zarr_res_level (int): The resolution level to use when reading the zarr files. Default is 0. 
+
+    Returns:
+    --------
+    None
+    """
+    print(f"Processing {colony} colony backdrops")
+    reader = load_data.get_dataset_original_file_reader(colony)
+    reader.set_resolution_level(zarr_res_level)
+
+    for timepoint_frame in range(reader.dims.T):
+        egfp = export_helper.load_raw_fov_image(colony, timepoint_frame, reader, channel="egfp")
+        egfp_flip = np.flip(egfp, axis=1)
+        egfp_mip = egfp_flip.max(axis=0).astype(dtype)
+
+        if downsample_factor is not None:
+            downsample_shape = (egfp_mip.shape[0] // downsample_factor, egfp_mip.shape[1] // downsample_factor)
+            egfp_mip = resize(egfp_mip, downsample_shape, anti_aliasing=True, preserve_range=True).astype(dtype)
+
+        egfp_mip = skex.rescale_intensity(image=egfp_mip, in_range=(110, 140), out_range=dtype).astype(dtype)
+
+        os.makedirs(figdir, exist_ok=True)
+        save_image = os.path.join(figdir, f'{timepoint_frame}.png')
+        two_d_writer.TwoDWriter.save(egfp_mip, uri=save_image)
+
+#%%