Invascular OCT Image Segmentation

Architecture

Architecture of U-net

Dependencies

• CentOS 7
• Python 3.6.5
• Python-tkinter
• Deploy the environment.

  pip install -r requirements.txt -i http://mirrors.aliyun.com/pypi/simple/ --trusted-host mirrors.aliyun.com

Usage

Data Preprocess

• Download source codes, run command:
  git clone https://github.com/zhyantao/image-segmentation.git
• Download TrainData.tar.gz to the directory image-segmentation/.
• Change directory:
  cd image-segmentation/src/
• Make original dataset, run command:
  sh make_dataset.sh
• Data preprocess, run command:
  python data_preprocess.py
• DIY your dataset, open DIY_dataset.sh and modify the number of images you wanna to train and valid, then run command:
  sh DIY_dataset.sh

Training model

python train.py

Test model

python test.py

Visualizing the result

python label_visualization.py

References

1. U-Net: Convolutional Networks for Biomedical Image Segmentation
2. Deep-Learning Based, Automated Segmentation of Macular Edema in Optical Coherence Tomography
3. SegNet: A Deep Convolutional Encoder-Decoder Architecture for Image Segmentation
4. Relationship Between a Systemic Inflammatory Marker, Plaque Inflammation, and Plaque Characteristics Determined by Intravascular Optical Coherence Tomography
5. preddy5/segnet

Results

For more details, see nohup.out

Visualizing Learning

tensorboard --logdir=../logs/