histopathology-inkRemoval

(working on removing explicit PATH setting...)

parts of codes are borrowed from CycleGAN for foreground ink removal

Requirements (maybe more than listed here, checkout requirements.txt):

• Linux/Unix only

• Keras with TensorFlow backend

• pyTorch 0.4

• CUDA 8.0/9.0

• Compiled Darknet for detection here or simply do

     -  ``git clone https://github.com/AlexeyAB/darknet.git``
     -  cd darknet 
     -  make (check for necessary settings [here](https://github.com/AlexeyAB/darknet))
     -  make a soft link ln -s $FULL_PATH/darknet $DIR_THIS_REPO/darknet/darknet
  

How to use it?

• Go to checkpoints folder and download test images and provided checkpoints
• Run: test_ink_removal.py in code folder (see parameters inside the code)

How to use the entire pipeline?

• Make sure you have a folder with image tiles
• Make sure you have all the checkpoints needed
• Go to scripts folder and edit writeTestDataResult.sh for PATHS
• If all goes well then you should be able to run the entire pipeline

Evaluation

Check the folder here

Test data

Download here

Note: Please change the data directory in scripts/writeTestDataResult.sh to '`pwd`/histo_ink_removal'

License

This software is covered by MIT License.

Cite

Ink removal from histopathology whole slide images by combining classification, detection and image generation models. Ali et al. 2019 IEEE 16th International Symposium on Biomedical Imaging (ISBI 2019)