U-Net Brain Tumor Segmentation

Last update: Jan 02, 2023

Overview

U-Net Brain Tumor Segmentation

🚀 ：Feb 2019 the data processing implementation in this repo is not the fastest way (code need update, contribution is welcome), you can use TensorFlow dataset API instead.

This repo show you how to train a U-Net for brain tumor segmentation. By default, you need to download the training set of BRATS 2017 dataset, which have 210 HGG and 75 LGG volumes, and put the data folder along with all scripts.

data
  -- Brats17TrainingData
  -- train_dev_all
model.py
train.py
...

About the data

Note that according to the license, user have to apply the dataset from BRAST, please do NOT contact me for the dataset. Many thanks.

Fig 1: Brain Image

Each volume have 4 scanning images: FLAIR、T1、T1c and T2.
Each volume have 4 segmentation labels:

Label 0: background
Label 1: necrotic and non-enhancing tumor
Label 2: edema 
Label 4: enhancing tumor

The prepare_data_with_valid.py split the training set into 2 folds for training and validating. By default, it will use only half of the data for the sake of training speed, if you want to use all data, just change DATA_SIZE = 'half' to all.

About the method

Network and Loss: In this experiment, as we use dice loss to train a network, one network only predict one labels (Label 1,2 or 4). We evaluate the performance using hard dice and IOU.
Data augmenation: Includes random left and right flip, rotation, shifting, shearing, zooming and the most important one -- Elastic trasnformation, see "Automatic Brain Tumor Detection and Segmentation Using U-Net Based Fully Convolutional Networks" for details.

Fig 2: Data augmentation

Start training

We train HGG and LGG together, as one network only have one task, set the task to all, necrotic, edema or enhance, "all" means learn to segment all tumors.

python train.py --task=all

Note that, if the loss stick on 1 at the beginning, it means the network doesn't converge to near-perfect accuracy, please try restart it.

Citation

If you find this project useful, we would be grateful if you cite the TensorLayer paper：

@article{tensorlayer2017,
author = {Dong, Hao and Supratak, Akara and Mai, Luo and Liu, Fangde and Oehmichen, Axel and Yu, Simiao and Guo, Yike},
journal = {ACM Multimedia},
title = {{TensorLayer: A Versatile Library for Efficient Deep Learning Development}},
url = {http://tensorlayer.org},
year = {2017}
}

Comments

TypeError: zoom_multi() got an unexpected keyword argument 'is_random'

Lossy conversion from float64 to uint8. Range [-0.18539370596408844, 2.158207416534424]. Convert image to uint8 prior to saving to suppress this warning. Traceback (most recent call last): File "train.py", line 250, in main(args.task) File "train.py", line 106, in main X[:,:,2,np.newaxis], X[:,:,3,np.newaxis], y])#[:,:,np.newaxis]]) File "train.py", line 26, in distort_imgs fill_mode='constant') TypeError: zoom_multi() got an unexpected keyword argument 'is_random'

opened by shenzeqi 8
MemoryError

@zsdonghao I am getting the memory error like this, What is the solution for this error?

Traceback (most recent call last): File "train.py", line 279, in main(args.task) File "train.py", line 78, in main y_test = (y_test > 0).astype(int) MemoryError

opened by PoonamZ 4
Error: Your CPU supports instructions that TensorFlow binary not compiled to use: AVX2

I am running run.py but gives error:

(base) G:>cd BraTS_2018_U-Net-master

(base) G:\BraTS_2018_U-Net-master>run.py [*] creates checkpoint ... [*] creates samples/all ... finished Brats18_2013_24_1 2019-06-15 22:05:45.959220: I tensorflow/core/platform/cpu_feature_guard.cc:141] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 Traceback (most recent call last): File "G:\BraTS_2018_U-Net-master\run.py", line 154, in

File "G:\BraTS_2018_U-Net-master\run.py", line 117, in main t_seg = tf.placeholder('float32', [1, nw, nh, 1], name='target_segment') NameError: name 'model' is not defined

opened by sapnii2 2
TypeError: __init__() got an unexpected keyword argument 'out_size'
After conv: Tensor("u_net/conv8/leaky_relu:0", shape=(5, 1, 1, 512), dtype=float32, device=/device:CPU:0) Traceback (most re cent call last): File "train.py", line 250, in main(args.task) File "train.py", line 121, in main net = model.u_net_bn(t_image, is_train=True, reuse=False, n_out=1) File "/home/achi/project/u-net-brain-tumor-master/model.py", line 179, in u_net_bn padding=pad, act=None, batch_size=batch_size, W_init=w_init, b_init=b_init, name='deconv7') File "/home/achi/anaconda3/lib/python3.6/site-packages/tensorlayer/decorators/deprecated_alias.py", line 24, in wrapper return f(*args, **kwargs) TypeError: init() got an unexpected keyword argument 'out_size'
opened by achintacsgit 1
Pre-trained model

I was wondering if you would share a pre-trained model. I would need to run inference-only, and training the model is taking longer than expected.

Thanks for sharing this project!

opened by luisremis 1
TypeError: zoom_multi() got an unexpected keyword argument 'is_random'

[TL] [!] checkpoint exists ... [TL] [!] samples/all exists ... Lossy conversion from float64 to uint8. Range [-0.19753389060497284, 2.826017379760742]. Convert image to uint8 prior to saving to suppress this warning.

TypeError Traceback (most recent call last) in 239 tl.files.save_npz(net.all_params, name=save_dir+'/u_net_{}.npz'.format(task), sess=sess) 240 --> 241 main(task='all') 242 243 ##if name == "main":

in main(task) 103 for i in range(10): 104 x_flair, x_t1, x_t1ce, x_t2, label = distort_imgs([X[:,:,0,np.newaxis], X[:,:,1,np.newaxis], --> 105 X[:,:,2,np.newaxis], X[:,:,3,np.newaxis], y])#[:,:,np.newaxis]]) 106 # print(x_flair.shape, x_t1.shape, x_t1ce.shape, x_t2.shape, label.shape) # (240, 240, 1) (240, 240, 1) (240, 240, 1) (240, 240, 1) (240, 240, 1) 107 X_dis = np.concatenate((x_flair, x_t1, x_t1ce, x_t2), axis=2)

in distort_imgs(data) 23 x1, x2, x3, x4, y = tl.prepro.zoom_multi([x1, x2, x3, x4, y], 24 zoom_range=[0.9, 1.1], is_random=True, ---> 25 fill_mode='constant') 26 return x1, x2, x3, x4, y 27

TypeError: zoom_multi() got an unexpected keyword argument 'is_random'

opened by BTapan 0
TensorFlow Implemetation

Do you have implementation of brain tumor segmentation code directly in tensorflow without using tensorlayer? If yes, can you share the same? Thank you.

opened by rupalkapdi 0
What is checkpoint?

When I run "python train.py" and then have a checkpoint folder is created. What function of checkpoint folder? Thank you

And I also have another question. When we had the picture, as follows. Is that the end result? I mean we can submit them to the Brast_2018 challenge?

Thank you very much.

opened by tphankr 0
Making sense

Novice here, i noticed the shape of the X_train arrays ended with 4. (240,240,4) Does each of those channel represent the type of the scan ( T1, t2, flair, t1ce ) ?

opened by guido-niku 1
Classification Layer - Activation & Shape?

Hi!

I went through this repository after reading your paper. Architecture on page 6, shows the final classification layer to produce feature maps of shape (240, 240, 2) which may indicate the use of a Softmax activation (not specified in the paper). On the contrary, model used in code has a classification layer of shape (240, 240, 1) using Sigmoid activation.

Kindly clarify this ambiguity.

opened by stalhabukhari 2