• For any query, contact us at
  • +91-9872993883
  • +91-8283824812
  • info@ris-ai.com
RIS AI - Thesis Expert
Most Viewed Articles
Object Detection | Start up Profit Prediction | RealTime Eye Blink Detection | House Budget Prediction | Human Detection and Counting | Pencil Sketch of Photo | Predict Next Word with Python | Hand Gesture Recognition | Handwritten Character Recognition
  • Blogs >

  • 3D Brain Tumor Segmentation and Identification in Python

3D Brain Tumor Segmentation and Identification in Python

After reading this notebook you guys learn, how 3D Brain Tumor Segmentation using UNET experiment is carried out using Brats Dataset. Here is the link Brats Dataset

In [ ]:
plotting.plot_stat_map('out/4.nii.gz')
Out[ ]:
<nilearn.plotting.displays.OrthoSlicer at 0x7f09bfc9e6a0>
3D Brain Tumor Segmentation and Identification in Python
In [ ]:
!pip install nilearn

Collecting nilearn
  Downloading https://files.pythonhosted.org/packages/b9/c2/f5f1bdd37a3da28b3b34305e4ba27cce468db6073998d62a38abd0e281da/nilearn-0.6.2-py3-none-any.whl (2.5MB)
     |████████████████████████████████| 2.5MB 7.3MB/s 
Requirement already satisfied: scikit-learn>=0.19 in /usr/local/lib/python3.6/dist-packages (from nilearn) (0.22.2.post1)
Requirement already satisfied: scipy>=0.19 in /usr/local/lib/python3.6/dist-packages (from nilearn) (1.4.1)
Requirement already satisfied: joblib>=0.11 in /usr/local/lib/python3.6/dist-packages (from nilearn) (0.15.1)
Requirement already satisfied: numpy>=1.11 in /usr/local/lib/python3.6/dist-packages (from nilearn) (1.18.5)
Requirement already satisfied: sklearn in /usr/local/lib/python3.6/dist-packages (from nilearn) (0.0)
Requirement already satisfied: nibabel>=2.0.2 in /usr/local/lib/python3.6/dist-packages (from nilearn) (3.0.2)
Installing collected packages: nilearn
Successfully installed nilearn-0.6.2
In [ ]:
!pip install SimpleITK

Collecting SimpleITK
  Downloading https://files.pythonhosted.org/packages/f8/d8/53338c34f71020725ffb3557846c80af96c29c03bc883551a2565aa68a7c/SimpleITK-1.2.4-cp36-cp36m-manylinux1_x86_64.whl (42.5MB)
     |████████████████████████████████| 42.5MB 70kB/s 
Installing collected packages: SimpleITK
Successfully installed SimpleITK-1.2.4
In [ ]:
!pwd

/content
In [ ]:
!pip show tensorflow-gpu

Name: tensorflow-gpu
Version: 1.15.2
Summary: TensorFlow is an open source machine learning framework for everyone.
Home-page: https://www.tensorflow.org/
Author: Google Inc.
Author-email: packages@tensorflow.org
License: Apache 2.0
Location: /usr/local/lib/python3.6/dist-packages
Requires: tensorflow-estimator, wrapt, grpcio, termcolor, tensorboard, opt-einsum, numpy, protobuf, absl-py, six, wheel, keras-preprocessing, gast, google-pasta, keras-applications, astor
Required-by:
In [2]:
import os
import glob
# os.environ['CUDA_VISIBLE_DEVICES'] = '-1'
from main_model import model_
import keras

config = dict()
config["image_shape"] = (32, 32, 32)  # This determines what shape the images will be cropped/resampled to.
config["patch_shape"] = None  # switch to None to train on the whole image

config["labels"] = (1)  # the label numbers on the input image
config["n_base_filters"] = 16
config["n_labels"] = 1
# config["all_modalities"] = ["t1"]
config["all_modalities"] = ["t1", "t1ce", "flair", "t2"]

config["training_modalities"] = config["all_modalities"]  # change this if you want to only use some of the modalities
config["nb_channels"] = len(config["training_modalities"])
if len(config["training_modalities"])>1:
    from all_data import write_data_to_file, open_data_file
else:
    from data import write_data_to_file, open_data_file


if "patch_shape" in config and config["patch_shape"] is not None:
    config["input_shape"] = tuple(list(config["patch_shape"]) + [config["nb_channels"]] )
else:
    config["input_shape"] = tuple( list(config["image_shape"]) + [config["nb_channels"]] )

config["truth_channel"] = config["nb_channels"]
config["deconvolution"] = True  #

config["batch_size"] = 1
config["validation_batch_size"] = 1
config["n_epochs"] = 500  # cutoff the training after this many epochs
config["patience"] = 10  # learning rate will be reduced after this many epochs if the validation loss is not improving
config["early_stop"] = 50  # training will be stopped after this many epochs without the validation loss improving
config["initial_learning_rate"] = 5e-4
config["learning_rate_drop"] = 0.5  # factor by which the learning rate will be reduced
config["validation_split"] = 0.8  # portion of the data that will be used for training
config["flip"] = False  # augments the data by randomly flipping an axis during
config["permute"] = True  # data shape must be a cube. Augments the data by permuting in various directions
config["distort"] = None  # switch to None if you want no distortion
config["augment"] = config["flip"] or config["distort"]
config["validation_patch_overlap"] = 0  # if > 0, during training, validation patches will be overlapping
config["training_patch_start_offset"] = (16, 16, 16)  # randomly offset the first patch index by up to this offset
config["skip_blank"] = True  # if True, then patches without any target will be skipped

config["data_file_train"] = os.path.abspath("brats_data_train.h5")
config["data_file_valid"] = os.path.abspath("brats_data_valid.h5")


config["overwrite"] = False

def fetch_training_data_files(return_subject_ids=False):
    training_data_files = list()
    subject_ids = list()
    for subject_dir in glob.glob(os.path.join(os.path.dirname(os.path.realpath('__file__')), "preprocessed", "*", "*")):
        print(subject_dir)
        # quit()
        subject_ids.append(os.path.basename(subject_dir))
        name_data = subject_dir.split('/')[-1]

        subject_files = list()
        for modality in config["training_modalities"] + ["seg"]:
            base_name = name_data + '_' + modality
            subject_files.append(os.path.join(subject_dir, base_name + ".nii.gz"))
        training_data_files.append(tuple(subject_files))
    if return_subject_ids:
        return training_data_files, subject_ids
    else:
        return training_data_files



def fetch_validate_data_files(return_subject_ids=False):
    training_data_files = list()
    subject_ids = list()
    for subject_dir in glob.glob(os.path.join(os.path.dirname(os.path.realpath('__file__')), "preproess_validate", "*", "*")):
        subject_ids.append(os.path.basename(subject_dir))
        name_data = subject_dir.split('/')[-1]
        subject_files = list()
        for modality in config["training_modalities"] + ["seg"]:
            base_name = name_data + '_' + modality
            subject_files.append(os.path.join(subject_dir, base_name + ".nii.gz"))
        training_data_files.append(tuple(subject_files))
    if return_subject_ids:
        return training_data_files, subject_ids
    else:
        return training_data_files
In [ ]:
# overwrite=config["overwrite"]
# convert input images into an hdf5 file
# if overwrite or not os.path.exists(config["data_file_train"]):
#     training_files, subject_ids = fetch_training_data_files(return_subject_ids=True)

    # write_data_to_file(training_files, config["data_file_train"], image_shape=config["image_shape"],
    #                     subject_ids=subject_ids)
data_file_opened = open_data_file(config["data_file_train"])
import numpy as np
train_data=np.array(data_file_opened.root.data)
# afine_data=np.array(data_file_opened.root.affine)

train_target=np.array(data_file_opened.root.truth)
del data_file_opened
In [ ]:
data_file_opened_test = open_data_file(config["data_file_valid"])
import numpy as np
test_data=np.array(data_file_opened_test.root.data)


test_target=np.array(data_file_opened_test.root.truth)
print(config["input_shape"])
del data_file_opened_test

(32, 32, 32, 4)
In [ ]:
from unet import unet_model_3d

model=unet_model_3d(input_shape=config["input_shape"],n_labels=1)

WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow_core/python/ops/resource_variable_ops.py:1630: calling BaseResourceVariable.__init__ (from tensorflow.python.ops.resource_variable_ops) with constraint is deprecated and will be removed in a future version.
Instructions for updating:
If using Keras pass *_constraint arguments to layers.
512
Model: "model_1"
__________________________________________________________________________________________________
Layer (type)                    Output Shape         Param #     Connected to                     
==================================================================================================
input_1 (InputLayer)            (None, 32, 32, 32, 4 0                                            
__________________________________________________________________________________________________
conv3d_1 (Conv3D)               (None, 32, 32, 32, 3 3488        input_1[0][0]                    
__________________________________________________________________________________________________
activation_1 (Activation)       (None, 32, 32, 32, 3 0           conv3d_1[0][0]                   
__________________________________________________________________________________________________
conv3d_2 (Conv3D)               (None, 32, 32, 32, 6 55360       activation_1[0][0]               
__________________________________________________________________________________________________
activation_2 (Activation)       (None, 32, 32, 32, 6 0           conv3d_2[0][0]                   
__________________________________________________________________________________________________
max_pooling3d_1 (MaxPooling3D)  (None, 16, 16, 16, 6 0           activation_2[0][0]               
__________________________________________________________________________________________________
conv3d_3 (Conv3D)               (None, 16, 16, 16, 6 110656      max_pooling3d_1[0][0]            
__________________________________________________________________________________________________
activation_3 (Activation)       (None, 16, 16, 16, 6 0           conv3d_3[0][0]                   
__________________________________________________________________________________________________
conv3d_4 (Conv3D)               (None, 16, 16, 16, 1 221312      activation_3[0][0]               
__________________________________________________________________________________________________
activation_4 (Activation)       (None, 16, 16, 16, 1 0           conv3d_4[0][0]                   
__________________________________________________________________________________________________
max_pooling3d_2 (MaxPooling3D)  (None, 8, 8, 8, 128) 0           activation_4[0][0]               
__________________________________________________________________________________________________
conv3d_5 (Conv3D)               (None, 8, 8, 8, 128) 442496      max_pooling3d_2[0][0]            
__________________________________________________________________________________________________
activation_5 (Activation)       (None, 8, 8, 8, 128) 0           conv3d_5[0][0]                   
__________________________________________________________________________________________________
conv3d_6 (Conv3D)               (None, 8, 8, 8, 256) 884992      activation_5[0][0]               
__________________________________________________________________________________________________
activation_6 (Activation)       (None, 8, 8, 8, 256) 0           conv3d_6[0][0]                   
__________________________________________________________________________________________________
max_pooling3d_3 (MaxPooling3D)  (None, 4, 4, 4, 256) 0           activation_6[0][0]               
__________________________________________________________________________________________________
conv3d_7 (Conv3D)               (None, 4, 4, 4, 256) 1769728     max_pooling3d_3[0][0]            
__________________________________________________________________________________________________
activation_7 (Activation)       (None, 4, 4, 4, 256) 0           conv3d_7[0][0]                   
__________________________________________________________________________________________________
conv3d_8 (Conv3D)               (None, 4, 4, 4, 512) 3539456     activation_7[0][0]               
__________________________________________________________________________________________________
activation_8 (Activation)       (None, 4, 4, 4, 512) 0           conv3d_8[0][0]                   
__________________________________________________________________________________________________
up_sampling3d_1 (UpSampling3D)  (None, 8, 8, 8, 512) 0           activation_8[0][0]               
__________________________________________________________________________________________________
concatenate_1 (Concatenate)     (None, 8, 8, 8, 768) 0           up_sampling3d_1[0][0]            
                                                                 activation_6[0][0]               
__________________________________________________________________________________________________
conv3d_9 (Conv3D)               (None, 8, 8, 8, 256) 5308672     concatenate_1[0][0]              
__________________________________________________________________________________________________
activation_9 (Activation)       (None, 8, 8, 8, 256) 0           conv3d_9[0][0]                   
__________________________________________________________________________________________________
conv3d_10 (Conv3D)              (None, 8, 8, 8, 256) 1769728     activation_9[0][0]               
__________________________________________________________________________________________________
activation_10 (Activation)      (None, 8, 8, 8, 256) 0           conv3d_10[0][0]                  
__________________________________________________________________________________________________
up_sampling3d_2 (UpSampling3D)  (None, 16, 16, 16, 2 0           activation_10[0][0]              
__________________________________________________________________________________________________
concatenate_2 (Concatenate)     (None, 16, 16, 16, 3 0           up_sampling3d_2[0][0]            
                                                                 activation_4[0][0]               
__________________________________________________________________________________________________
conv3d_11 (Conv3D)              (None, 16, 16, 16, 1 1327232     concatenate_2[0][0]              
__________________________________________________________________________________________________
activation_11 (Activation)      (None, 16, 16, 16, 1 0           conv3d_11[0][0]                  
__________________________________________________________________________________________________
conv3d_12 (Conv3D)              (None, 16, 16, 16, 1 442496      activation_11[0][0]              
__________________________________________________________________________________________________
activation_12 (Activation)      (None, 16, 16, 16, 1 0           conv3d_12[0][0]                  
__________________________________________________________________________________________________
up_sampling3d_3 (UpSampling3D)  (None, 32, 32, 32, 1 0           activation_12[0][0]              
__________________________________________________________________________________________________
concatenate_3 (Concatenate)     (None, 32, 32, 32, 1 0           up_sampling3d_3[0][0]            
                                                                 activation_2[0][0]               
__________________________________________________________________________________________________
conv3d_13 (Conv3D)              (None, 32, 32, 32, 6 331840      concatenate_3[0][0]              
__________________________________________________________________________________________________
activation_13 (Activation)      (None, 32, 32, 32, 6 0           conv3d_13[0][0]                  
__________________________________________________________________________________________________
conv3d_14 (Conv3D)              (None, 32, 32, 32, 6 110656      activation_13[0][0]              
__________________________________________________________________________________________________
activation_14 (Activation)      (None, 32, 32, 32, 6 0           conv3d_14[0][0]                  
__________________________________________________________________________________________________
conv3d_15 (Conv3D)              (None, 32, 32, 32, 1 65          activation_14[0][0]              
__________________________________________________________________________________________________
activation_15 (Activation)      (None, 32, 32, 32, 1 0           conv3d_15[0][0]                  
==================================================================================================
Total params: 16,318,177
Trainable params: 16,318,177
Non-trainable params: 0
__________________________________________________________________________________________________
None
In [ ]:
from main_model import model_
from metrics import weighted_dice_coefficient_loss
from keras.optimizers import Adam


# model = model_(input_shape=config["input_shape"], depth=3, n_labels=1,
#                           initial_learning_rate=0.00001,
#                           n_base_filters=4)
# model.compile(optimizer=Adam(lr=0.1), loss=weighted_dice_coefficient_loss)
    # loss = "binary_crossentropy", metrics = ["accuracy"]

# print(model.summary())

from unet import unet_model_3d

model=unet_model_3d(input_shape=config["input_shape"],n_labels=1)
learning_rate_reduction = keras.callbacks.ReduceLROnPlateau(monitor='val_loss',
                                                                patience=10,
                                                                verbose=1,
                                                                factor=0.5,
                                                                min_lr=0.000001)

history=model.fit(train_data,train_target,validation_data=(test_data,test_target),batch_size=1,verbose=1,initial_epoch=1,nb_epoch=10,callbacks=[learning_rate_reduction])
# model.save('trained_big.h5')

512
Model: "model_2"
__________________________________________________________________________________________________
Layer (type)                    Output Shape         Param #     Connected to                     
==================================================================================================
input_2 (InputLayer)            (None, 32, 32, 32, 4 0                                            
__________________________________________________________________________________________________
conv3d_16 (Conv3D)              (None, 32, 32, 32, 3 3488        input_2[0][0]                    
__________________________________________________________________________________________________
activation_16 (Activation)      (None, 32, 32, 32, 3 0           conv3d_16[0][0]                  
__________________________________________________________________________________________________
conv3d_17 (Conv3D)              (None, 32, 32, 32, 6 55360       activation_16[0][0]              
__________________________________________________________________________________________________
activation_17 (Activation)      (None, 32, 32, 32, 6 0           conv3d_17[0][0]                  
__________________________________________________________________________________________________
max_pooling3d_4 (MaxPooling3D)  (None, 16, 16, 16, 6 0           activation_17[0][0]              
__________________________________________________________________________________________________
conv3d_18 (Conv3D)              (None, 16, 16, 16, 6 110656      max_pooling3d_4[0][0]            
__________________________________________________________________________________________________
activation_18 (Activation)      (None, 16, 16, 16, 6 0           conv3d_18[0][0]                  
__________________________________________________________________________________________________
conv3d_19 (Conv3D)              (None, 16, 16, 16, 1 221312      activation_18[0][0]              
__________________________________________________________________________________________________
activation_19 (Activation)      (None, 16, 16, 16, 1 0           conv3d_19[0][0]                  
__________________________________________________________________________________________________
max_pooling3d_5 (MaxPooling3D)  (None, 8, 8, 8, 128) 0           activation_19[0][0]              
__________________________________________________________________________________________________
conv3d_20 (Conv3D)              (None, 8, 8, 8, 128) 442496      max_pooling3d_5[0][0]            
__________________________________________________________________________________________________
activation_20 (Activation)      (None, 8, 8, 8, 128) 0           conv3d_20[0][0]                  
__________________________________________________________________________________________________
conv3d_21 (Conv3D)              (None, 8, 8, 8, 256) 884992      activation_20[0][0]              
__________________________________________________________________________________________________
activation_21 (Activation)      (None, 8, 8, 8, 256) 0           conv3d_21[0][0]                  
__________________________________________________________________________________________________
max_pooling3d_6 (MaxPooling3D)  (None, 4, 4, 4, 256) 0           activation_21[0][0]              
__________________________________________________________________________________________________
conv3d_22 (Conv3D)              (None, 4, 4, 4, 256) 1769728     max_pooling3d_6[0][0]            
__________________________________________________________________________________________________
activation_22 (Activation)      (None, 4, 4, 4, 256) 0           conv3d_22[0][0]                  
__________________________________________________________________________________________________
conv3d_23 (Conv3D)              (None, 4, 4, 4, 512) 3539456     activation_22[0][0]              
__________________________________________________________________________________________________
activation_23 (Activation)      (None, 4, 4, 4, 512) 0           conv3d_23[0][0]                  
__________________________________________________________________________________________________
up_sampling3d_4 (UpSampling3D)  (None, 8, 8, 8, 512) 0           activation_23[0][0]              
__________________________________________________________________________________________________
concatenate_4 (Concatenate)     (None, 8, 8, 8, 768) 0           up_sampling3d_4[0][0]            
                                                                 activation_21[0][0]              
__________________________________________________________________________________________________
conv3d_24 (Conv3D)              (None, 8, 8, 8, 256) 5308672     concatenate_4[0][0]              
__________________________________________________________________________________________________
activation_24 (Activation)      (None, 8, 8, 8, 256) 0           conv3d_24[0][0]                  
__________________________________________________________________________________________________
conv3d_25 (Conv3D)              (None, 8, 8, 8, 256) 1769728     activation_24[0][0]              
__________________________________________________________________________________________________
activation_25 (Activation)      (None, 8, 8, 8, 256) 0           conv3d_25[0][0]                  
__________________________________________________________________________________________________
up_sampling3d_5 (UpSampling3D)  (None, 16, 16, 16, 2 0           activation_25[0][0]              
__________________________________________________________________________________________________
concatenate_5 (Concatenate)     (None, 16, 16, 16, 3 0           up_sampling3d_5[0][0]            
                                                                 activation_19[0][0]              
__________________________________________________________________________________________________
conv3d_26 (Conv3D)              (None, 16, 16, 16, 1 1327232     concatenate_5[0][0]              
__________________________________________________________________________________________________
activation_26 (Activation)      (None, 16, 16, 16, 1 0           conv3d_26[0][0]                  
__________________________________________________________________________________________________
conv3d_27 (Conv3D)              (None, 16, 16, 16, 1 442496      activation_26[0][0]              
__________________________________________________________________________________________________
activation_27 (Activation)      (None, 16, 16, 16, 1 0           conv3d_27[0][0]                  
__________________________________________________________________________________________________
up_sampling3d_6 (UpSampling3D)  (None, 32, 32, 32, 1 0           activation_27[0][0]              
__________________________________________________________________________________________________
concatenate_6 (Concatenate)     (None, 32, 32, 32, 1 0           up_sampling3d_6[0][0]            
                                                                 activation_17[0][0]              
__________________________________________________________________________________________________
conv3d_28 (Conv3D)              (None, 32, 32, 32, 6 331840      concatenate_6[0][0]              
__________________________________________________________________________________________________
activation_28 (Activation)      (None, 32, 32, 32, 6 0           conv3d_28[0][0]                  
__________________________________________________________________________________________________
conv3d_29 (Conv3D)              (None, 32, 32, 32, 6 110656      activation_28[0][0]              
__________________________________________________________________________________________________
activation_29 (Activation)      (None, 32, 32, 32, 6 0           conv3d_29[0][0]                  
__________________________________________________________________________________________________
conv3d_30 (Conv3D)              (None, 32, 32, 32, 1 65          activation_29[0][0]              
__________________________________________________________________________________________________
activation_30 (Activation)      (None, 32, 32, 32, 1 0           conv3d_30[0][0]                  
==================================================================================================
Total params: 16,318,177
Trainable params: 16,318,177
Non-trainable params: 0
__________________________________________________________________________________________________
None

/usr/local/lib/python3.6/dist-packages/ipykernel_launcher.py:23: UserWarning: The `nb_epoch` argument in `fit` has been renamed `epochs`.

WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:422: The name tf.global_variables is deprecated. Please use tf.compat.v1.global_variables instead.

Train on 285 samples, validate on 98 samples
Epoch 2/10
285/285 [==============================] - 89s 312ms/step - loss: -0.2476 - dice_coefficient: 0.2476 - val_loss: -0.3466 - val_dice_coefficient: 0.3466
Epoch 3/10
285/285 [==============================] - 82s 287ms/step - loss: -0.3296 - dice_coefficient: 0.3296 - val_loss: -0.3569 - val_dice_coefficient: 0.3569
Epoch 4/10
285/285 [==============================] - 82s 287ms/step - loss: -0.3443 - dice_coefficient: 0.3443 - val_loss: -0.3861 - val_dice_coefficient: 0.3861
Epoch 5/10
285/285 [==============================] - 81s 286ms/step - loss: -0.3598 - dice_coefficient: 0.3598 - val_loss: -0.4107 - val_dice_coefficient: 0.4107
Epoch 6/10
285/285 [==============================] - 81s 286ms/step - loss: -0.4070 - dice_coefficient: 0.4070 - val_loss: -0.5036 - val_dice_coefficient: 0.5036
Epoch 7/10
285/285 [==============================] - 81s 285ms/step - loss: -0.5312 - dice_coefficient: 0.5312 - val_loss: -0.6344 - val_dice_coefficient: 0.6344
Epoch 8/10
285/285 [==============================] - 81s 284ms/step - loss: -0.6028 - dice_coefficient: 0.6028 - val_loss: -0.6938 - val_dice_coefficient: 0.6938
Epoch 9/10
285/285 [==============================] - 81s 284ms/step - loss: -0.6598 - dice_coefficient: 0.6598 - val_loss: -0.6725 - val_dice_coefficient: 0.6725
Epoch 10/10
285/285 [==============================] - 81s 283ms/step - loss: -0.7233 - dice_coefficient: 0.7233 - val_loss: -0.7054 - val_dice_coefficient: 0.7054
In [ ]:
model.evaluate(test_data,test_target)

98/98 [==============================] - 11s 113ms/step
Out[ ]:
[-0.753577535249749, 0.8017180562019348]
In [ ]:
# history=model.fit(train_data,train_target,validation_data=(test_data,test_target),batch_size=1,verbose=1,initial_epoch=10,nb_epoch=50,callbacks=[learning_rate_reduction])

Train on 285 samples, validate on 98 samples
Epoch 11/50
  1/285 [..............................] - ETA: 38s - loss: -0.3105
/usr/local/lib/python3.6/dist-packages/ipykernel_launcher.py:1: UserWarning: The `nb_epoch` argument in `fit` has been renamed `epochs`.
  """Entry point for launching an IPython kernel.

285/285 [==============================] - 21s 75ms/step - loss: -0.3671 - val_loss: -0.2720
Epoch 12/50
285/285 [==============================] - 22s 76ms/step - loss: -0.3743 - val_loss: -0.2380
Epoch 13/50
285/285 [==============================] - 21s 75ms/step - loss: -0.3748 - val_loss: -0.1618
Epoch 14/50
285/285 [==============================] - 21s 75ms/step - loss: -0.3805 - val_loss: -0.2399
Epoch 15/50
285/285 [==============================] - 21s 75ms/step - loss: -0.3875 - val_loss: -0.1963
Epoch 16/50
285/285 [==============================] - 21s 75ms/step - loss: -0.3833 - val_loss: -0.1810
Epoch 17/50
285/285 [==============================] - 21s 75ms/step - loss: -0.3952 - val_loss: -0.1649
Epoch 18/50
285/285 [==============================] - 21s 75ms/step - loss: -0.3990 - val_loss: -0.1757
Epoch 19/50
285/285 [==============================] - 21s 75ms/step - loss: -0.3999 - val_loss: -0.1627
Epoch 20/50
285/285 [==============================] - 21s 75ms/step - loss: -0.4045 - val_loss: -0.2240
Epoch 21/50
285/285 [==============================] - 21s 75ms/step - loss: -0.4072 - val_loss: -0.1508

Epoch 00021: ReduceLROnPlateau reducing learning rate to 0.05000000074505806.
Epoch 22/50
285/285 [==============================] - 21s 75ms/step - loss: -0.4212 - val_loss: -0.2022
Epoch 23/50
285/285 [==============================] - 21s 75ms/step - loss: -0.4282 - val_loss: -0.1755
Epoch 24/50
285/285 [==============================] - 21s 75ms/step - loss: -0.4294 - val_loss: -0.1609
Epoch 25/50
285/285 [==============================] - 21s 75ms/step - loss: -0.4351 - val_loss: -0.1766
Epoch 26/50
285/285 [==============================] - 21s 75ms/step - loss: -0.4373 - val_loss: -0.1744
Epoch 27/50
285/285 [==============================] - 21s 75ms/step - loss: -0.4375 - val_loss: -0.1598
Epoch 28/50
285/285 [==============================] - 21s 75ms/step - loss: -0.4428 - val_loss: -0.1651
Epoch 29/50
285/285 [==============================] - 21s 75ms/step - loss: -0.4446 - val_loss: -0.1542
Epoch 30/50
285/285 [==============================] - 21s 75ms/step - loss: -0.4511 - val_loss: -0.1899
Epoch 31/50
285/285 [==============================] - 21s 75ms/step - loss: -0.5053 - val_loss: -0.1737

Epoch 00031: ReduceLROnPlateau reducing learning rate to 0.02500000037252903.
Epoch 32/50
285/285 [==============================] - 21s 75ms/step - loss: -0.5425 - val_loss: -0.1700
Epoch 33/50
285/285 [==============================] - 21s 75ms/step - loss: -0.5483 - val_loss: -0.1592
Epoch 34/50
285/285 [==============================] - 21s 75ms/step - loss: -0.5633 - val_loss: -0.1618
Epoch 35/50
285/285 [==============================] - 21s 75ms/step - loss: -0.5580 - val_loss: -0.1673
Epoch 36/50
285/285 [==============================] - 21s 75ms/step - loss: -0.5702 - val_loss: -0.1743
Epoch 37/50
285/285 [==============================] - 21s 75ms/step - loss: -0.5727 - val_loss: -0.1746
Epoch 38/50
285/285 [==============================] - 21s 75ms/step - loss: -0.5762 - val_loss: -0.1982
Epoch 39/50
285/285 [==============================] - 21s 75ms/step - loss: -0.5862 - val_loss: -0.1746
Epoch 40/50
285/285 [==============================] - 21s 75ms/step - loss: -0.5938 - val_loss: -0.1728
Epoch 41/50
285/285 [==============================] - 21s 75ms/step - loss: -0.5935 - val_loss: -0.1590

Epoch 00041: ReduceLROnPlateau reducing learning rate to 0.012500000186264515.
Epoch 42/50
285/285 [==============================] - 21s 74ms/step - loss: -0.6060 - val_loss: -0.1616
Epoch 43/50
285/285 [==============================] - 21s 75ms/step - loss: -0.6131 - val_loss: -0.1680
Epoch 44/50
285/285 [==============================] - 21s 75ms/step - loss: -0.6143 - val_loss: -0.1605
Epoch 45/50
285/285 [==============================] - 21s 75ms/step - loss: -0.6227 - val_loss: -0.1719
Epoch 46/50
285/285 [==============================] - 21s 75ms/step - loss: -0.6228 - val_loss: -0.1675
Epoch 47/50
285/285 [==============================] - 21s 75ms/step - loss: -0.6266 - val_loss: -0.1652
Epoch 48/50
285/285 [==============================] - 21s 75ms/step - loss: -0.6276 - val_loss: -0.1752
Epoch 49/50
285/285 [==============================] - 21s 75ms/step - loss: -0.6259 - val_loss: -0.1678
Epoch 50/50
285/285 [==============================] - 21s 75ms/step - loss: -0.6308 - val_loss: -0.1855
In [ ]:
import matplotlib.pyplot as plt
def plot_model_history(model_history):
    fig, axs = plt.subplots(1,2,figsize=(15,5))
    # summarize history for accuracy
    
    # summarize history for loss
    axs[0].plot(range(1,len(model_history.history['loss'])+1),model_history.history['loss'])
    axs[0].plot(range(1,len(model_history.history['val_loss'])+1),model_history.history['val_loss'])
    axs[0].set_title('Model Loss')
    axs[0].set_ylabel('Loss')
    axs[0].set_xlabel('Epoch')
    axs[0].set_xticks(np.arange(1,len(model_history.history['loss'])+1),len(model_history.history['loss'])/10)
    axs[0].legend(['train', 'val'], loc='best')

    axs[1].plot(range(1,len(model_history.history['dice_coefficient'])+1),model_history.history['dice_coefficient'])
    axs[1].plot(range(1,len(model_history.history['val_dice_coefficient'])+1),model_history.history['val_dice_coefficient'])
    axs[1].set_title('Model Dice Coefficient')
    axs[1].set_ylabel('Dice Coefficient')
    axs[1].set_xlabel('Epoch')
    axs[1].set_xticks(np.arange(1,len(model_history.history['loss'])+1),len(model_history.history['loss'])/10)
    axs[1].legend(['train Dice Coefficient', 'val Dice Coefficient'], loc='best')
    plt.show()
plot_model_history(history)

/usr/local/lib/python3.6/dist-packages/ipykernel_launcher.py:12: MatplotlibDeprecationWarning: Passing the minor parameter of set_xticks() positionally is deprecated since Matplotlib 3.2; the parameter will become keyword-only two minor releases later.
  if sys.path[0] == '':
/usr/local/lib/python3.6/dist-packages/ipykernel_launcher.py:20: MatplotlibDeprecationWarning: Passing the minor parameter of set_xticks() positionally is deprecated since Matplotlib 3.2; the parameter will become keyword-only two minor releases later.
3d Brain Tumor segmentation Model Graph
In [ ]:
sample_test=test_data[0]
sample_test.shape
resample_data=sample_test[np.newaxis]
resample_data.shape
response=model.predict(resample_data)
out_resp=response[0]
out_l=out_resp[:,:,:,0]*4

real=test_target[0][:,:,:,0]
In [ ]:
test_data.shape
Out[ ]:
(98, 32, 32, 32, 4)
In [ ]:
sample_test=test_data[0:9]
In [ ]:
sample_test.shape
Out[ ]:
(9, 32, 32, 32, 4)
In [ ]:
type(sample_test)
Out[ ]:
numpy.ndarray
In [ ]:
resample_data=sample_test[np.newaxis]
In [ ]:
resample_data.shape
Out[ ]:
(1, 32, 32, 32, 4)
In [ ]:
response=model.predict(sample_test)
In [ ]:
response.shape
Out[ ]:
(9, 32, 32, 32, 1)
In [ ]:
out_resp=response[4]
In [ ]:
out_resp.shape
Out[ ]:
(32, 32, 32, 1)
In [ ]:
out_l=out_resp[:,:,:,0]
In [ ]:
out_l.shape
Out[ ]:
(32, 32, 32)
In [ ]:
test_target.shape
Out[ ]:
(98, 32, 32, 32, 1)
In [ ]:
real=test_target[4]
real.shape
Out[ ]:
(32, 32, 32, 1)
In [ ]:
real_l=real[:,:,:,0]
In [ ]:
real_l.shape
Out[ ]:
(32, 32, 32)
In [ ]:
import nibabel
import numpy as np
import random
import os
import SimpleITK as sitk
import pickle
from scipy import ndimage
def save_to_nii(im, filename, outdir="", mode="image", system="sitk"):

    if system == "sitk":
        if mode == 'label':
            img = sitk.GetImageFromArray(im.astype(np.uint8))
        else:
            img = sitk.GetImageFromArray(im.astype(np.float32))
        if not os.path.exists("./{}".format(outdir)):
            os.mkdir("./{}".format(outdir))
        sitk.WriteImage(img, "./{}/{}.nii.gz".format(outdir, filename))
    else:
        img = np.rot90(im, k=2, axes= (1,2))
        OUTPUT_AFFINE = np.array(
                [[0, 0, 1, 0],
                [0, 1, 0, 0],
                [1, 0, 0, 0],
                [0, 0, 0, 1]])
        if mode == 'label':
            img = nibabel.Nifti1Image(img.astype(np.uint8), OUTPUT_AFFINE)
        else:
            img = nibabel.Nifti1Image(img.astype(np.float32), OUTPUT_AFFINE)
        if not os.path.exists("./{}".format(outdir)):
            os.mkdir("./{}".format(outdir))
        nibabel.save(img, "./{}/{}.nii.gz".format(outdir, filename))
In [ ]:
sample_test2=test_data[7]
resample_data2=sample_test[np.newaxis]
response2=model.predict(resample_data)
out_resp2=response2[0]
out_l2=out_resp2[:,:,:,0]

real2=test_target[4][:,:,:,0]
In [ ]:
plotting.plot_stat_map('out/3.nii.gz')
Out[ ]:
<nilearn.plotting.displays.OrthoSlicer at 0x7f09c22acb38>
3d Brain Tumor segmentation in Python

Search Article

For getting more closer to code contact us . Our Expert Data Scientist will get in touch with you at hourly price of $51
Entire implementation is done on Tensorflow V2

Resources You Will Ever Need

Popular Searches