Cleanup pc preprocessing

docs/PREPROCESSING.md

@@ -37,15 +37,17 @@ Preprocessing is divided into three steps that use two different virtual environ
 
 # Punctate structures: Generate pointclouds
 
-Edit the data paths in the following file to match the location of the outputs of the alignment, masking, and registration step, then run it.
+Use the preprocessed data manifest generated via the alignment, masking, and registration steps from image as input to the pointcloud generation step
 
 ```
 src
 └── br
     └── data
         └── preprocessing
             └── pc_preprocessing
-                └── punctate_cyto.py <- Point cloud sampling from raw images for punctate structures here
+                └── pcna.py <- Point cloud sampling from raw images for DNA replication foci dataset here
+                └── punctate_nuc.py <- Point cloud sampling from raw images of nuclear structures from the WTC-11 hIPS single cell image dataset here
+                └── punctate_cyto.py <- Point cloud sampling from raw images of cytoplasmic structures from the WTC-11 hIPS single cell image dataset here
 ```
 
 # Polymorphic structures: Generate SDFs

src/br/analysis/visualize_pointclouds.py

@@ -80,7 +80,7 @@ def main(args):
     for _, this_image in orig_image_df.iterrows():
         cell_id = this_image["CellId"]
         if not strat:
-            strat_val = this_image['structure_name']
+            strat_val = this_image["structure_name"]
 
         if args.dataset_name == "pcna":
             points_all, _, img, center = compute_labels_pcna(this_image, False)
@@ -108,10 +108,12 @@ def main(args):
         if mem_ind is not None:
             img_mem = img[mem_ind]
 
-        if (args.dataset_name == 'other_punctate') and (strat_val in ["CETN2", "RAB5A", "SLC25A17"]):
-            img_raw = np.where(img_mem, img_raw, 0) # mask by mem/nuc seg
+        if (args.dataset_name == "other_punctate") and (
+            strat_val in ["CETN2", "RAB5A", "SLC25A17"]
+        ):
+            img_raw = np.where(img_mem, img_raw, 0)  # mask by mem/nuc seg
         else:
-            img_raw = np.where(img_nuc, img_raw, 0) # mask by mem/nuc seg
+            img_raw = np.where(img_nuc, img_raw, 0)  # mask by mem/nuc seg
 
         # Sample sparse point cloud and get images
         probs2 = points_all["s"].values
@@ -168,7 +170,7 @@ def main(args):
             center_slice=center_slice,
         )
         ax_array[2].set_title("Sampling sparse PC")
-        print(f'Saving {name}.png')
+        print(f"Saving {name}.png")
         fig.savefig(Path(args.save_path) / Path(f"{name}.png"), bbox_inches="tight", dpi=300)
 
 

src/br/data/preprocessing/pc_preprocessing/pcna.py

@@ -1,4 +1,6 @@
+import argparse
 from multiprocessing import Pool
+from pathlib import Path
 
 import numpy as np
 import pandas as pd
@@ -64,26 +66,34 @@ def compute_labels(row, save=True):
 
     cell_id = str(row["CellId"])
 
-    save_path = path_prefix + cell_id + ".ply"
+    save_path = Path(path_prefix) / Path(cell_id + ".ply")
 
     new_cents = new_cents.astype(float)
 
     cloud = PyntCloud(new_cents)
-    cloud.to_file(save_path)
+    cloud.to_file(str(save_path))
 
 
 def get_center_of_mass(img):
     center_of_mass = np.mean(np.stack(np.where(img > 0)), axis=1)
     return np.floor(center_of_mass + 0.5).astype(int)
 
 
-if __name__ == "__main__":
-    df = pd.read_csv(PCNA_SINGLE_CELL_PATH)
+def main(args):
+
+    # make save path directory
+    Path(args.save_path).mkdir(parents=True, exist_ok=True)
+
+    df = pd.read_parquet(args.preprocessed_manifest)
+
+    if args.global_path:
+        df["registered_path"] = df["registered_path"].apply(lambda x: args.global_path + x)
 
-    path_prefix = SAVE_LOCATION
+    global path_prefix
+    path_prefix = args.save_path
 
     all_rows = []
-    for ind, row in tqdm(df.iterrows(), total=len(df)):
+    for _, row in tqdm(df.iterrows(), total=len(df)):
         all_rows.append(row)
 
     with Pool(40) as p:
@@ -97,3 +107,31 @@ def get_center_of_mass(img):
                 desc="compute_everything",
             )
         )
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(
+        description="Script for computing point clouds for PCNA dataset"
+    )
+    parser.add_argument("--save_path", type=str, required=True, help="Path to save results.")
+    parser.add_argument(
+        "--global_path",
+        type=str,
+        default=None,
+        required=False,
+        help="Path to append to relative paths in preprocessed manifest",
+    )
+    parser.add_argument(
+        "--preprocessed_manifest",
+        type=str,
+        required=True,
+        help="Path to processed single cell image manifest.",
+    )
+    args = parser.parse_args()
+    main(args)
+
+    """
+    Example run: 
+
+    python src/br/data/preprocessing/pc_preprocessing/pcna --save_path "./make_pcs_test" --preprocessed_manifest "./subpackages/image_preprocessing/tmp_output_pcna/processed/manifest.parquet" --global_path "./subpackages/image_preprocessing/"
+    """

src/br/data/preprocessing/pc_preprocessing/punctate_cyto.py

@@ -1,5 +1,7 @@
+import argparse
 import warnings
 from multiprocessing import Pool
+from pathlib import Path
 
 import numpy as np
 import pandas as pd
@@ -96,25 +98,34 @@ def compute_labels(row, save=True):
 
     cell_id = str(row["CellId"])
 
-    save_path = path_prefix + cell_id + ".ply"
+    save_path = Path(path_prefix) / Path(cell_id + ".ply")
 
     new_cents = new_cents.astype(float)
     cloud = PyntCloud(new_cents)
-    cloud.to_file(save_path)
+    cloud.to_file(str(save_path))
 
 
 def get_center_of_mass(img):
     center_of_mass = np.mean(np.stack(np.where(img > 0)), axis=1)
     return np.floor(center_of_mass + 0.5).astype(int)
 
 
-if __name__ == "__main__":
-    df = pd.read_parquet(SINGLE_CELL_IMAGE_PATH)
+def main(args):
+
+    # make save path directory
+    Path(args.save_path).mkdir(parents=True, exist_ok=True)
+
+    df = pd.read_parquet(args.preprocessed_manifest)
+    df = df.loc[df["structure_name"].isin(SKEW_EXP_DICT.keys())]
 
-    path_prefix = SAVE_LOCATION
+    if args.global_path:
+        df["registered_path"] = df["registered_path"].apply(lambda x: args.global_path + x)
+
+    global path_prefix
+    path_prefix = args.save_path
 
     all_rows = []
-    for ind, row in tqdm(df.iterrows(), total=len(df)):
+    for _, row in tqdm(df.iterrows(), total=len(df)):
         all_rows.append(row)
 
     with Pool(40) as p:
@@ -128,3 +139,31 @@ def get_center_of_mass(img):
                 desc="compute_everything",
             )
         )
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(
+        description="Script for computing point clouds for cytoplasmic structures from WTC-11 hIPS single cell image dataset"
+    )
+    parser.add_argument("--save_path", type=str, required=True, help="Path to save results.")
+    parser.add_argument(
+        "--global_path",
+        type=str,
+        default=None,
+        required=False,
+        help="Path to append to relative paths in preprocessed manifest",
+    )
+    parser.add_argument(
+        "--preprocessed_manifest",
+        type=str,
+        required=True,
+        help="Path to processed single cell image manifest.",
+    )
+    args = parser.parse_args()
+    main(args)
+
+    """
+    Example run: 
+
+    python src/br/data/preprocessing/pc_preprocessing/punctate_cyto.py --save_path "./make_pcs_test" --preprocessed_manifest "./subpackages/image_preprocessing/tmp_output_variance/processed/manifest.parquet" --global_path "./subpackages/image_preprocessing/"
+    """

src/br/data/preprocessing/pc_preprocessing/punctate_nuc.py

@@ -1,10 +1,16 @@
+import argparse
+from multiprocessing import Pool
+from pathlib import Path
+
 import numpy as np
 import pandas as pd
 from pyntcloud import PyntCloud
 from scipy.ndimage import binary_dilation
 from skimage.io import imread
 from tqdm import tqdm
 
+STRUCTS = ["HIST1H2BJ", "NUP153", "SMC1A", "SON"]
+
 
 def compute_labels(row, save=True):
     path = row["registered_path"]
@@ -64,31 +70,35 @@ def compute_labels(row, save=True):
 
     cell_id = str(row["CellId"])
 
-    save_path = path_prefix + cell_id + ".ply"
+    save_path = Path(path_prefix) / Path(cell_id + ".ply")
 
     new_cents = new_cents.astype(float)
     cloud = PyntCloud(new_cents)
-    cloud.to_file(save_path)
+    cloud.to_file(str(save_path))
 
 
 def get_center_of_mass(img):
     center_of_mass = np.mean(np.stack(np.where(img > 0)), axis=1)
     return np.floor(center_of_mass + 0.5).astype(int)
 
 
-if __name__ == "__main__":
-    df = pd.read_parquet(SINGLE_CELL_IMAGE_PATH)
+def main(args):
+
+    # make save path directory
+    Path(args.save_path).mkdir(parents=True, exist_ok=True)
+
+    df = pd.read_parquet(args.preprocessed_manifest)
+    df = df.loc[df["structure_name"].isin(STRUCTS)]
 
-    path_prefix = SAVE_LOCATION
+    if args.global_path:
+        df["registered_path"] = df["registered_path"].apply(lambda x: args.global_path + x)
+
+    global path_prefix
+    path_prefix = args.save_path
 
     all_rows = []
-    for ind, row in tqdm(df.iterrows(), total=len(df)):
+    for _, row in tqdm(df.iterrows(), total=len(df)):
         all_rows.append(row)
-        # if str(row['CellId']) == '660844':
-        #     print('yes')
-        #     compute_labels(row)
-
-    from multiprocessing import Pool
 
     with Pool(40) as p:
         _ = tuple(
@@ -101,3 +111,31 @@ def get_center_of_mass(img):
                 desc="compute_everything",
             )
         )
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(
+        description="Script for computing point clouds for nuclear structures from WTC-11 hIPS single cell image dataset"
+    )
+    parser.add_argument("--save_path", type=str, required=True, help="Path to save results.")
+    parser.add_argument(
+        "--global_path",
+        type=str,
+        default=None,
+        required=False,
+        help="Path to append to relative paths in preprocessed manifest",
+    )
+    parser.add_argument(
+        "--preprocessed_manifest",
+        type=str,
+        required=True,
+        help="Path to processed single cell image manifest.",
+    )
+    args = parser.parse_args()
+    main(args)
+
+    """
+    Example run: 
+
+    python src/br/data/preprocessing/pc_preprocessing/punctate_nuc.py --save_path "./make_pcs_test" --preprocessed_manifest "./subpackages/image_preprocessing/tmp_output_variance/processed/manifest.parquet" --global_path "./subpackages/image_preprocessing/"
+    """