Added nuclei filtering to labeled images

Dockerfile

@@ -16,6 +16,7 @@ RUN apt-get update -y && apt-get install -y \
     libpng-dev \
     default-jdk \
     libtiff-dev \
+    git \
     && rm -rf /var/lib/apt/lists/*
 
 # rJava configuration

Python/fabgal/calculate_CTF.py

@@ -1,6 +1,9 @@
 from pathlib import Path
 import pandas as pd
 from re import sub
+from skimage import io
+from skimage.color import label2rgb
+import numpy as np
 import logging
 from .config import FABgalConfig
 from .helpers import choose_threshold
@@ -39,12 +42,15 @@ def calculate_CTF(cfg: FABgalConfig):
 
         # BiaPy results (prior or newly generated)
         if cfg.Biapy_run is not None:
+            biapyout = Path(cfg.out_path) / f"Results_{cfg.experiment_name}" / cfg.Biapy_run / "BiaPy_output"
             biapystatsfile = Path(cfg.out_path) / f"Results_{cfg.experiment_name}" / cfg.Biapy_run / "BiaPy_results.tsv"
         else:
+            biapyout = results_dir / "BiaPy_output"
             biapystatsfile = results_dir / "BiaPy_results.tsv"
     else:
         # Newly generated results for both if is_rerun is false
         bgalquantfile = results_dir / "Raw_Bgal_results.tsv"
+        biapyout = results_dir / "BiaPy_output"
         biapystatsfile = results_dir / "BiaPy_results.tsv"
 
     ### Check if B-gal and BiaPy results file exist ###
@@ -122,7 +128,7 @@ def calculate_CTF(cfg: FABgalConfig):
         nucleidf = pd.read_table(biapystatsfile)
         bgaldf = pd.read_table(bgalquantfile)
 
-        # Filter nuclei below area threshold
+        ##### Filter nuclei below area threshold #####
         nucleidf_pxarea = pd.merge(nucleidf,bgaldf[['File','PxArea']],how='inner',on='File')
 
         ## If cfg.nuclei_thr is None, then choose interactively the nuclei thr
@@ -133,6 +139,34 @@ def calculate_CTF(cfg: FABgalConfig):
         else:
             nucleidf_pxarea['nucl_thr_pixel'] = cfg.nuclei_thr / nucleidf_pxarea.PxArea
             nucleidf_filt = nucleidf_pxarea[nucleidf_pxarea.area > nucleidf_pxarea.nucl_thr_pixel]
+        ## If cfg.nuclei_thr is None, then choose interactively the nuclei thr
+        if cfg.nuclei_thr is None:
+            cfg.nuclei_thr = choose_threshold(nucleidf_pxarea)
+            nucleidf_pxarea['nucl_thr_pixel'] = cfg.nuclei_thr / nucleidf_pxarea.PxArea
+            nucleidf_filt = nucleidf_pxarea[nucleidf_pxarea.area > nucleidf_pxarea.nucl_thr_pixel]
+        else:
+            nucleidf_pxarea['nucl_thr_pixel'] = cfg.nuclei_thr / nucleidf_pxarea.PxArea
+            nucleidf_filt = nucleidf_pxarea[nucleidf_pxarea.area > nucleidf_pxarea.nucl_thr_pixel]
+
+        ## Filter BiaPy mask results ##
+        original_masks = biapyout / "original_masks"
+
+        filtered_masks = results_dir / "filtered_masks"
+        filtered_masks.mkdir(exist_ok=True)
+
+        if cfg.keep_masks:
+            for inf in original_masks.glob("*.tif"):
+                img = io.imread(inf)
+                #### Filter by area ####
+                label, area = np.unique(img, return_counts = True)
+                labs_to_remove = label[area < cfg.nuclei_thr / nucleidf_pxarea['PxArea'].median()]
+                img[np.isin(img,labs_to_remove)] = 0
+
+                ## Save mask as RGB ##
+                img = label2rgb(img)*255
+                img = img.astype(np.uint8)
+                io.imsave(filtered_masks / f"{inf.stem}_mask.png",img, check_contrast = False)
+
 
         # Count nuclei per image file 
         nucleitot = nucleidf_filt['File'].value_counts()

Python/fabgal/run_biapy.py

@@ -3,11 +3,9 @@
 import logging
 from biapy import BiaPy
 import pandas as pd
-from skimage import io
-from skimage.color import label2rgb
-import numpy as np
 import shutil
 from .config import FABgalConfig
+from .helpers import choose_threshold
 
 #### Log options ####
 
@@ -54,7 +52,7 @@ def run_biapy(cfg: FABgalConfig):
     ## Define Nuclei stats output file
     biapystatsfile = results_dir / "BiaPy_results.tsv"
 
-    #### Process nuclei stats ####
+    ######## Process nuclei stats ########
 
     # List of csv with filename column
     biapy_csv = list(biapyout.glob("*full_stats.csv"))
@@ -68,23 +66,23 @@ def run_biapy(cfg: FABgalConfig):
     # Concatenate dfs
     result = pd.concat(dfs, ignore_index=True)
     result = result.drop(columns=["conditions"])
+
+    # Save concat data
     result.to_csv(str(biapystatsfile), sep="\t", encoding="utf-8")
 
     # Delete original csv
     for inf in biapy_csv:
         inf.unlink()
 
-    ########### Arrange intermediate files and images ###########
+    ########## Arrange intermediate files and images ###########
 
-    #### Process output mask images ####
+    ### Process output mask images ####
+    masks_out = results_dir / "BiaPy_output" / "original_masks"
+    masks_out.mkdir(exist_ok=True)
 
     if cfg.keep_masks:
-        for inf in biapyout.glob("*.tif"):
-            img = io.imread(inf)
-            img = label2rgb(img)*255
-            img = img.astype(np.uint8)
-            io.imsave(results_dir / "BiaPy_output" / f"{inf.stem}_mask.png",img, check_contrast = False)
-            inf.unlink()
+        for file in biapyout.glob("*.tif"):
+            file.replace(masks_out / file.name)
 
     #### Save subtracted images from BiaPy input (if generated) ####