/NuClick

Interactive Segmentation of Nucleus

Primary LanguagePythonGNU General Public License v3.0GPL-3.0

NuClick

Clicks inside the nuclei or draw a scribble inside glands to obtain segmentation. This is Keras implementation of NuClick (link to paper )

alt textalt text alt text

Dataset

The datasets released as part of NuClick paper are:

  • Images of white blood cells (WBC) in blood sample images with their segmentation masks (Download)
  • Datasets of lymphocyte segmentation in Immunohistochemistry (IHC) images (Download)

Requirements

Keras==2.2.4 h5py==2.9.0 opencv_python_headless==4.1.0.25 pandas==0.25.1 matplotlib==3.1.1 six==1.12.0 tensorflow_gpu==1.13.1 scipy==1.3.0 numpy==1.17.0 albumentations==0.3.1 scikit_image==0.15.0 Pillow==8.0.1 skimage==0.0 tensorflow==2.4.0

Inference:

Download weights for nucleus segmentation from here and save it inside weights folder:

  • Set application in the config.py
  • Run test.py to use the interface.

To apply it on the existing detection points:

  • Save points in the .mat format: first column x
  • Define directories in the config.py: mat_path = '' , images_path = '' and save_path = ''
  • run test_all_images