initial release

.gitattributes

@@ -33,3 +33,14 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+*.nii filter=lfs diff=lfs merge=lfs -text
+examples/ex_1/BraTS20_Training_001_flair.nii filter=lfs diff=lfs merge=lfs -text
+examples/ex_1/BraTS20_Training_001_seg.nii filter=lfs diff=lfs merge=lfs -text
+examples/ex_1/BraTS20_Training_001_t1.nii filter=lfs diff=lfs merge=lfs -text
+examples/ex_1/BraTS20_Training_001_t1ce.nii filter=lfs diff=lfs merge=lfs -text
+examples/ex_1/BraTS20_Training_001_t2.nii filter=lfs diff=lfs merge=lfs -text
+examples/ex_2/BraTS20_Training_002_flair.nii filter=lfs diff=lfs merge=lfs -text
+examples/ex_2/BraTS20_Training_002_t2.nii filter=lfs diff=lfs merge=lfs -text
+examples/ex_2/BraTS20_Training_002_t1ce.nii filter=lfs diff=lfs merge=lfs -text
+examples/ex_2/BraTS20_Training_002_t1.nii filter=lfs diff=lfs merge=lfs -text
+examples/ex_2/BraTS20_Training_002_seg.nii filter=lfs diff=lfs merge=lfs -text

.idea/.gitignore

@@ -0,0 +1,8 @@
+# Default ignored files
+/shelf/
+/workspace.xml
+# Editor-based HTTP Client requests
+/httpRequests/
+# Datasource local storage ignored files
+/dataSources/
+/dataSources.local.xml

.idea/Brain-Tumor-Segmentation.iml

@@ -0,0 +1,8 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<module type="PYTHON_MODULE" version="4">
+  <component name="NewModuleRootManager">
+    <content url="file://$MODULE_DIR$" />
+    <orderEntry type="inheritedJdk" />
+    <orderEntry type="sourceFolder" forTests="false" />
+  </component>
+</module>

.idea/inspectionProfiles/Project_Default.xml

@@ -0,0 +1,27 @@
+<component name="InspectionProjectProfileManager">
+  <profile version="1.0">
+    <option name="myName" value="Project Default" />
+    <inspection_tool class="DuplicatedCode" enabled="true" level="WEAK WARNING" enabled_by_default="true">
+      <Languages>
+        <language minSize="89" name="Python" />
+      </Languages>
+    </inspection_tool>
+    <inspection_tool class="Eslint" enabled="true" level="WARNING" enabled_by_default="true" />
+    <inspection_tool class="PyPep8NamingInspection" enabled="true" level="WEAK WARNING" enabled_by_default="true">
+      <option name="ignoredErrors">
+        <list>
+          <option value="N806" />
+        </list>
+      </option>
+    </inspection_tool>
+    <inspection_tool class="PyUnresolvedReferencesInspection" enabled="true" level="WARNING" enabled_by_default="true">
+      <option name="ignoredIdentifiers">
+        <list>
+          <option value="map.__getitem__" />
+          <option value="cv2.drawMarker" />
+          <option value="cv2.putText" />
+        </list>
+      </option>
+    </inspection_tool>
+  </profile>
+</component>

.idea/inspectionProfiles/profiles_settings.xml

@@ -0,0 +1,6 @@
+<component name="InspectionProjectProfileManager">
+  <settings>
+    <option name="USE_PROJECT_PROFILE" value="false" />
+    <version value="1.0" />
+  </settings>
+</component>

.idea/modules.xml

@@ -0,0 +1,8 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<project version="4">
+  <component name="ProjectModuleManager">
+    <modules>
+      <module fileurl="file://$PROJECT_DIR$/.idea/Brain-Tumor-Segmentation.iml" filepath="$PROJECT_DIR$/.idea/Brain-Tumor-Segmentation.iml" />
+    </modules>
+  </component>
+</project>

.idea/vcs.xml

@@ -0,0 +1,6 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<project version="4">
+  <component name="VcsDirectoryMappings">
+    <mapping directory="$PROJECT_DIR$" vcs="Git" />
+  </component>
+</project>

app.py

@@ -0,0 +1,148 @@
+import numpy as np
+import cv2
+import nibabel as nib
+from PIL import Image
+import io
+import matplotlib.pyplot as plt
+import gradio as gr
+
+from huggingface_hub import from_pretrained_keras
+
+model = from_pretrained_keras("duzduran/NeuroNest3D")
+# Constants
+IMG_SIZE = 128
+VOLUME_SLICES = 100
+VOLUME_START_AT = 22
+SEGMENT_CLASSES = ['NOT tumor', 'ENHANCING', 'CORE', 'WHOLE']
+
+
+def predictByPath(flair, ce):
+    X = np.empty((VOLUME_SLICES, IMG_SIZE, IMG_SIZE, 2))
+
+    for j in range(VOLUME_SLICES):
+        X[j, :, :, 0] = cv2.resize(flair[:, :, j + VOLUME_START_AT], (IMG_SIZE, IMG_SIZE))
+        X[j, :, :, 1] = cv2.resize(ce[:, :, j + VOLUME_START_AT], (IMG_SIZE, IMG_SIZE))
+
+    # Normalize and make predictions
+    X_normalized = X / np.max(X)
+    return model.predict(X_normalized, verbose=1)
+
+
+def create_subplot_image(origImage, gt, predictions, slice_index, start_at, img_size):
+    plt.figure(figsize=(18, 10))
+    f, axarr = plt.subplots(1, 6, figsize=(18, 10))
+
+    for i in range(6):
+        axarr[i].imshow(cv2.resize(origImage[:, :, slice_index + start_at], (img_size, img_size)), cmap="gray",
+                        interpolation='none')
+
+    # Original image flair
+    axarr[0].title.set_text('Original image flair')
+
+    # Ground truth
+    curr_gt = cv2.resize(gt[:, :, slice_index + start_at], (img_size, img_size), interpolation=cv2.INTER_NEAREST)
+    axarr[1].imshow(curr_gt, cmap="Reds", interpolation='none', alpha=0.3)
+    axarr[1].title.set_text('Ground truth')
+
+    # All classes
+    axarr[2].imshow(predictions[slice_index, :, :, 1:4], cmap="Reds", interpolation='none', alpha=0.3)
+    axarr[2].title.set_text('All classes')
+
+    SEGMENT_CLASSES
+    # Class-specific predictions
+    for i in range(1, 4):  # Adjusted to loop over the available prediction classes
+        axarr[i + 2].imshow(predictions[slice_index, :, :, i], cmap="OrRd", interpolation='none', alpha=0.3)
+        axarr[i + 2].title.set_text(f'{SEGMENT_CLASSES[i]} predicted')
+
+    # Convert plot to image
+    buf = io.BytesIO()
+    plt.savefig(buf, format='png')
+    plt.close(f)
+    buf.seek(0)
+    img = Image.open(buf)
+    return img
+
+
+examples = {
+    "Example 1": {"flair": "examples/ex_1/BraTS20_Training_001_flair.nii",
+                  "t1ce": "examples/ex_1/BraTS20_Training_001_t1ce.nii",
+                  "seg": "examples/ex_1/BraTS20_Training_001_seg.nii"},
+
+    "Example 2": {"flair": "examples/ex_2/BraTS20_Training_002_flair.nii",
+                  "t1ce": "examples/ex_2/BraTS20_Training_002_t1ce.nii",
+                  "seg": "examples/ex_2/BraTS20_Training_002_seg.nii"},
+}
+
+
+def automatic_process(example_key):
+    paths = examples[example_key]
+    print(paths["flair"])
+
+    flair = nib.load(paths["flair"]).get_fdata()
+    t1ce = nib.load(paths["t1ce"]).get_fdata()
+    seg = nib.load(paths["seg"]).get_fdata()
+
+    # Default slice index
+    slice_index = 50
+
+    return process_and_display_direct(flair, t1ce, seg, slice_index)
+
+
+def process_and_display_direct(flair_data, t1ce_data, seg_data, slice_index):
+    flair = np.array(flair_data)
+    t1ce = np.array(t1ce_data)
+    seg = np.array(seg_data)
+
+    p = predictByPath(flair, t1ce)
+
+    # Create the subplot image
+    subplot_img = create_subplot_image(flair, seg, p, slice_index, VOLUME_START_AT, IMG_SIZE)
+
+    return subplot_img
+
+
+def process_and_display(flair_file, t1ce_file, seg_file, slice_index):
+    if not flair_file or not t1ce_file or not seg_file:
+        return None  # Ensure all files are uploaded
+
+    flair = nib.load(flair_file.name).get_fdata()
+    t1ce = nib.load(t1ce_file.name).get_fdata()
+    gt = nib.load(seg_file.name).get_fdata()
+
+    p = predictByPath(flair, t1ce)
+
+    # Create the subplot image
+    subplot_img = create_subplot_image(flair, gt, p, slice_index, VOLUME_START_AT, IMG_SIZE)
+
+    return subplot_img
+
+
+# Gradio Interface
+with gr.Blocks() as demo:
+    with gr.Row():
+        flair_input = gr.File(label="Upload Flair NIfTI File")
+        t1ce_input = gr.File(label="Upload T1ce NIfTI File")
+        seg_input = gr.File(label="Upload Seg NIfTI File")
+        slice_input = gr.Slider(minimum=0, maximum=VOLUME_SLICES - 1, label="Slice Index")
+        #eval_class_input = gr.Dropdown(choices=list(range(len(SEGMENT_CLASSES))), label="Select Class")
+        submit_button = gr.Button("Submit")
+
+    with gr.Row():
+        example_selector = gr.Dropdown(list(examples.keys()), label="Select Example")
+        auto_button = gr.Button("Load Example")
+
+    output_image = gr.Image(label="Visualization")
+
+    submit_button.click(
+        process_and_display,
+        inputs=[flair_input, t1ce_input, seg_input, slice_input],
+        outputs=output_image
+    )
+
+    auto_button.click(
+        automatic_process,
+        inputs=[example_selector],
+        outputs=output_image
+    )
+
+demo.launch()

environment.yml

@@ -0,0 +1,25 @@
+name: my_ml_env
+channels:
+  - conda-forge
+  - defaults
+dependencies:
+  - python=3.8  
+  - pip
+  - gradio
+  - numpy
+  - pandas
+  - seaborn
+  - matplotlib
+  - scikit-image
+  - pillow
+  - tensorflow
+  - keras
+  - scikit-learn
+  - opencv
+  - nibabel
+  - nilearn
+  - tensorflow-estimator
+  - h5py
+  - pip:
+    - git+https://github.com/miykael/gif_your_nifti  
+

examples/ex_1/BraTS20_Training_001_flair.nii

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1671dfab75fbaaede29989011cabcef12b7825c72a315f293441b698a8e8d39c
+size 17858880

examples/ex_1/BraTS20_Training_001_seg.nii

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:56f39a503ebc4df200a5b9872b1ace83d92fa775d55b6ce301f12d2f63507180
+size 8930976

examples/ex_1/BraTS20_Training_001_t1.nii

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6440b711f433a7071379088fc61ad9e0ff01ae5f551a7e5f006135c5adcf0f83
+size 17858880

examples/ex_1/BraTS20_Training_001_t1ce.nii

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ed3732f26987226748147562cc64e58958564a6a7e47611b726773fd4e057b99
+size 17858880

examples/ex_1/BraTS20_Training_001_t2.nii

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0918a2aed1e8aac1a802bf899940f9032b0e61bae70b35c9fa93d35e7e27c569
+size 17858880

examples/ex_2/BraTS20_Training_002_flair.nii

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f51b5715db978afb246dc00b9e2950bcdb7a48ff0933f0a1b8f4ebd0d08c8613
+size 17858896

examples/ex_2/BraTS20_Training_002_seg.nii

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e5476e303ce247b623ea9fb25fc08833831560c71bc9bc5b93e637f0482fdb5
+size 8930976

examples/ex_2/BraTS20_Training_002_t1.nii

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e0cc02bda2df29e8100459fe730b9cd5bf3350ce5d79a53ea17c39d8b9378c50
+size 17858880

examples/ex_2/BraTS20_Training_002_t1ce.nii

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cd9402472cbc67d9c0f65102fc198c23649c509deceb418fc2c69cbd91ddbd78
+size 17858880

examples/ex_2/BraTS20_Training_002_t2.nii

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:741f3823e914e21efeff9d1795e84bef7ee3acf701d50f758b32e658ab094cda
+size 17858880

loss.py

@@ -0,0 +1,57 @@
+import keras.backend as K
+
+# dice loss as defined above for 4 classes
+def dice_coef(y_true, y_pred, smooth=1.0):
+    class_num = 4
+    for i in range(class_num):
+        y_true_f = K.flatten(y_true[:,:,:,i])
+        y_pred_f = K.flatten(y_pred[:,:,:,i])
+        intersection = K.sum(y_true_f * y_pred_f)
+        loss = ((2. * intersection + smooth) / (K.sum(y_true_f) + K.sum(y_pred_f) + smooth))
+   #     K.print_tensor(loss, message='loss value for class {} : '.format(SEGMENT_CLASSES[i]))
+        if i == 0:
+            total_loss = loss
+        else:
+            total_loss = total_loss + loss
+    total_loss = total_loss / class_num
+#    K.print_tensor(total_loss, message=' total dice coef: ')
+    return total_loss
+
+
+
+# define per class evaluation of dice coef
+# inspired by https://github.com/keras-team/keras/issues/9395
+def dice_coef_necrotic(y_true, y_pred, epsilon=1e-6):
+    intersection = K.sum(K.abs(y_true[:,:,:,1] * y_pred[:,:,:,1]))
+    return (2. * intersection) / (K.sum(K.square(y_true[:,:,:,1])) + K.sum(K.square(y_pred[:,:,:,1])) + epsilon)
+
+def dice_coef_edema(y_true, y_pred, epsilon=1e-6):
+    intersection = K.sum(K.abs(y_true[:,:,:,2] * y_pred[:,:,:,2]))
+    return (2. * intersection) / (K.sum(K.square(y_true[:,:,:,2])) + K.sum(K.square(y_pred[:,:,:,2])) + epsilon)
+
+def dice_coef_enhancing(y_true, y_pred, epsilon=1e-6):
+    intersection = K.sum(K.abs(y_true[:,:,:,3] * y_pred[:,:,:,3]))
+    return (2. * intersection) / (K.sum(K.square(y_true[:,:,:,3])) + K.sum(K.square(y_pred[:,:,:,3])) + epsilon)
+
+
+
+# Computing Precision
+def precision(y_true, y_pred):
+        true_positives = K.sum(K.round(K.clip(y_true * y_pred, 0, 1)))
+        predicted_positives = K.sum(K.round(K.clip(y_pred, 0, 1)))
+        precision = true_positives / (predicted_positives + K.epsilon())
+        return precision
+
+
+# Computing Sensitivity
+def sensitivity(y_true, y_pred):
+    true_positives = K.sum(K.round(K.clip(y_true * y_pred, 0, 1)))
+    possible_positives = K.sum(K.round(K.clip(y_true, 0, 1)))
+    return true_positives / (possible_positives + K.epsilon())
+
+
+# Computing Specificity
+def specificity(y_true, y_pred):
+    true_negatives = K.sum(K.round(K.clip((1-y_true) * (1-y_pred), 0, 1)))
+    possible_negatives = K.sum(K.round(K.clip(1-y_true, 0, 1)))
+    return true_negatives / (possible_negatives + K.epsilon())