Dev: refactorize

fastMONAI/__init__.py

@@ -1 +1 @@
-__version__ = "0.3.0"
+__version__ = "0.3.1"

fastMONAI/_modidx.py

@@ -189,12 +189,10 @@
                                                                                        'fastMONAI/vision_inference.py'),
                                             'fastMONAI.vision_inference._to_original_orientation': ( 'vision_inference.html#_to_original_orientation',
                                                                                                      'fastMONAI/vision_inference.py'),
-                                            'fastMONAI.vision_inference.find_similar_size_labels': ( 'vision_inference.html#find_similar_size_labels',
-                                                                                                     'fastMONAI/vision_inference.py'),
                                             'fastMONAI.vision_inference.inference': ( 'vision_inference.html#inference',
                                                                                       'fastMONAI/vision_inference.py'),
-                                            'fastMONAI.vision_inference.pred_postprocess': ( 'vision_inference.html#pred_postprocess',
-                                                                                             'fastMONAI/vision_inference.py')},
+                                            'fastMONAI.vision_inference.refine_binary_pred_mask': ( 'vision_inference.html#refine_binary_pred_mask',
+                                                                                                    'fastMONAI/vision_inference.py')},
             'fastMONAI.vision_loss': { 'fastMONAI.vision_loss.CustomLoss': ( 'vision_loss_functions.html#customloss',
                                                                              'fastMONAI/vision_loss.py'),
                                        'fastMONAI.vision_loss.CustomLoss.__call__': ( 'vision_loss_functions.html#customloss.__call__',

fastMONAI/vision_inference.py

@@ -1,7 +1,7 @@
 # AUTOGENERATED! DO NOT EDIT! File to edit: ../nbs/06_vision_inference.ipynb.
 
 # %% auto 0
-__all__ = ['inference', 'pred_postprocess', 'find_similar_size_labels']
+__all__ = ['inference', 'refine_binary_pred_mask']
 
 # %% ../nbs/06_vision_inference.ipynb 1
 import numpy as np
@@ -56,27 +56,36 @@ def inference(learn_inf, reorder, resample, fn:(Path,str)='', save_path:(str,Pat
     return org_img
 
 # %% ../nbs/06_vision_inference.ipynb 7
-def pred_postprocess(pred_mask, remove_size=10437, percentage=0.2): 
-    '''Remove small objects from predicted mask. (TODO:refactorize)'''
-    small_objects = remove_size*percentage    
-    labeled_mask, ncomponents = label(pred_mask)
-    labeled_mask = remove_small_objects(labeled_mask, min_size=small_objects)
-    
-    return np.where(labeled_mask>0, 1., 0.)
+def refine_binary_pred_mask(
+        pred_mask,
+        remove_size: (int, float) = None,
+        percentage: float = 0.2,
+        verbose: bool = False
+):
+    """Removes small objects from the predicted binary mask.
 
-# %% ../nbs/06_vision_inference.ipynb 8
-def find_similar_size_labels(labeled_mask, size_threshold=0.8):
-    """
-    Find labels of components in a labeled mask that are of similar size 
-    to the largest component.
+    Args:
+        pred_mask: The predicted mask from which small objects are to be removed.
+        remove_size: The size under which objects are considered 'small'.
+        percentage: The percentage of the remove_size to be used as threshold. 
+            Defaults to 0.2.
+        verbose: If True, print the number of components. Defaults to False.
+
+    Returns:
+        The processed mask with small objects removed.
     """
-
-    sizes = np.bincount(labeled_mask.ravel())
-    max_label = sizes[1:].argmax() + 1
-    threshold_size = size_threshold * sizes[max_label]
-    similar_size_labels = [
-        label for label, size in enumerate(sizes[1:], start=1)
-        if size >= threshold_size
-    ]
+    labeled_mask, n_components = label(pred_mask)
+
+    if verbose:
+        print(n_components)
+
+    if remove_size is None:
+        sizes = np.bincount(labeled_mask.ravel())
+        max_label = sizes[1:].argmax() + 1
+        remove_size = sizes[max_label]
+
+    small_objects_threshold = remove_size * percentage
+    processed_mask = remove_small_objects(
+        labeled_mask, min_size=small_objects_threshold)
 
-    return max_label, similar_size_labels
+    return np.where(processed_mask > 0, 1., 0.)

nbs/06_vision_inference.ipynb

@@ -114,43 +114,44 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d6000a96-56fb-4966-b4f4-83dcb51684e9",
+   "id": "a64a3407-4b97-4b1c-933c-d4a316dbff94",
    "metadata": {},
    "outputs": [],
    "source": [
     "#| export\n",
-    "def pred_postprocess(pred_mask, remove_size=10437, percentage=0.2): \n",
-    "    '''Remove small objects from predicted mask. (TODO:refactorize)'''\n",
-    "    small_objects = remove_size*percentage    \n",
-    "    labeled_mask, ncomponents = label(pred_mask)\n",
-    "    labeled_mask = remove_small_objects(labeled_mask, min_size=small_objects)\n",
-    "    \n",
-    "    return np.where(labeled_mask>0, 1., 0.)"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "id": "f491d6e0-c8d6-460c-9a13-17561aa64ac4",
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "#| export\n",
-    "def find_similar_size_labels(labeled_mask, size_threshold=0.8):\n",
-    "    \"\"\"\n",
-    "    Find labels of components in a labeled mask that are of similar size \n",
-    "    to the largest component.\n",
+    "def refine_binary_pred_mask(\n",
+    "        pred_mask,\n",
+    "        remove_size: (int, float) = None,\n",
+    "        percentage: float = 0.2,\n",
+    "        verbose: bool = False\n",
+    "):\n",
+    "    \"\"\"Removes small objects from the predicted binary mask.\n",
+    "\n",
+    "    Args:\n",
+    "        pred_mask: The predicted mask from which small objects are to be removed.\n",
+    "        remove_size: The size under which objects are considered 'small'.\n",
+    "        percentage: The percentage of the remove_size to be used as threshold. \n",
+    "            Defaults to 0.2.\n",
+    "        verbose: If True, print the number of components. Defaults to False.\n",
+    "\n",
+    "    Returns:\n",
+    "        The processed mask with small objects removed.\n",
     "    \"\"\"\n",
-    "    \n",
-    "    sizes = np.bincount(labeled_mask.ravel())\n",
-    "    max_label = sizes[1:].argmax() + 1\n",
-    "    threshold_size = size_threshold * sizes[max_label]\n",
-    "    similar_size_labels = [\n",
-    "        label for label, size in enumerate(sizes[1:], start=1)\n",
-    "        if size >= threshold_size\n",
-    "    ]\n",
+    "    labeled_mask, n_components = label(pred_mask)\n",
+    "\n",
+    "    if verbose:\n",
+    "        print(n_components)\n",
+    "\n",
+    "    if remove_size is None:\n",
+    "        sizes = np.bincount(labeled_mask.ravel())\n",
+    "        max_label = sizes[1:].argmax() + 1\n",
+    "        remove_size = sizes[max_label]\n",
+    "\n",
+    "    small_objects_threshold = remove_size * percentage\n",
+    "    processed_mask = remove_small_objects(\n",
+    "        labeled_mask, min_size=small_objects_threshold)\n",
     "\n",
-    "    return max_label, similar_size_labels"
+    "    return np.where(processed_mask > 0, 1., 0.)"
    ]
   }
  ],

research/endometrical_cancer/inference_script.py

@@ -7,7 +7,6 @@
 
 from fastMONAI.vision_all import *
 from huggingface_hub import snapshot_download
-from scipy.ndimage import label
 
 def find_similar_size_labels(labeled_mask, size_threshold=0.8):
     """
@@ -51,15 +50,8 @@ def find_similar_size_labels(labeled_mask, size_threshold=0.8):
 #pred_items
 org_img, input_img, org_size = med_img_reader(img_path, reorder=reorder, resample=resample, only_tensor=False)
 
-    #Predict with ensemble
 mask_data = inference(learner, reorder=reorder, resample=resample, org_img=org_img, input_img=input_img, org_size=org_size).data 
-labeled_mask, ncomponents = label(mask_data.numpy())
-if ncomponents > 1: 
-    max_label, similar_size_labels = find_similar_size_labels(labeled_mask)
-    if len(similar_size_labels) == 1:
-        mask_data = torch.Tensor(np.where(labeled_mask == max_label, 1, 0))
-    else: 
-        print(save_path)
+mask_data = refine_binary_pred_mask(mask_data, percentage=0.8)
 
 if "".join(org_img.orientation) == 'LSA':        
     mask_data = mask_data.permute(0,1,3,2)

research/spine/inference_script.py

@@ -5,7 +5,6 @@
 import argparse
 
 from fastMONAI.vision_all import *
-from IPython.display import clear_output
 from huggingface_hub import snapshot_download
 
 # Parse command line arguments
@@ -48,7 +47,7 @@
 mask_data = torch.where(mask_data > 0.5, 1., 0.)
 
 # Apply postprocessing to remove small objects from the binary mask
-mask_data = torch.Tensor(pred_postprocess(mask_data))
+mask_data = refine_binary_pred_mask(mask_data, remove_size=10437, percentage=0.2)
 
 # Set the data of the mask object to the processed mask data
 mask.set_data(mask_data)

settings.ini

@@ -5,7 +5,7 @@
 ### Python Library ###
 lib_name = fastMONAI
 min_python = 3.7
-version = 0.3.0
+version = 0.3.1
 ### OPTIONAL ###
 
 requirements = fastai==2.7.12 monai==1.2.0 torchio==0.18.91 xlrd>=1.2.0 scikit-image==0.19.3 huggingface-hub gdown