DMMN 2 CLAM

This repository contains the code to allow DMMN-ovary segmentation approach to be integrated into the CLAM framework.

Problem

...

Solution

...

Download and installation

Both CLAM and DMMN-Ovary GitHub repositories should be downloaded from their original project repos.

dmmn2clam repo can be downloaded here using the following command from command line:

git clone https://github.com/micGuerr/dmmn2clam.git

Let's assume CLAM's repo is downloaded in <fullPath_to_CLAM_repo>, DMMN-Ovary repo is downloaded in <fullPath_to_DMMN-ovary_repo> and dmmn2clam repo is downloaded in <fullPath_to_dmmn2clam_repo>.

Installation

1. Configure both CLAM and DMMN-Ovary GitHub repositories in such a way they can be used in their standard configuration.

2. Run the following commands from command line:

Link the DMMN-ovary files into the DMMN-ovary directory:

ln -s <fullPath_to_dmmn2clam_repo>m/slidereader_coords_dmmn2clam.py <fullPath_to_DMMN-ovary_repo>/slidereader_coords_dmmn2clam.py

ln -s <fullPath_to_dmmn2clam_repo>/inference_dmmn2clam.py <fullPath_to_DMMN-ovary_repo>/inference_dmmn2clam.py

ln -s <fullPath_to_dmmn2clam_repo>/extract_tissue_dmmn2clam.py <fullPath_to_DMMN-ovary_repo>/extract_tissue_dmmn2clam.py

Link CLAM's files into CLAM directory:

ln -s <fullPath_to_dmmn2clam_repo>/create_patches_fp_dmmn2clam.py <fullPath_to_CLAM_repo>/create_patches_fp_dmmn2clam.py

ln -s <fullPath_to_dmmn2clam_repo>/WholeSlideImage_dmmn2clam.py <fullPath_to_CLAM_repo>/wsi_core/WholeSlideImage_dmmn2clam.py

ln -s <fullPath_to_dmmn2clam_repo>/batch_process_utils_dmmn2clam.py <fullPath_to_CLAM_repo>/wsi_core/batch_process_utils_dmmn2clam.py

Note that in Windows the link can be constructed from prompt window, opened as admin, with mklink command, e.g.:

mklink<fullPath_to_DMMN-ovary_repo>/inference_dmmn2clam.py <fullPath_to_dmmn2clam_repo>/inference_dmmn2clam.py

Usage

Below an example of the commands needed to integrate the DMMN-ovary segmentation into CLAM (here I assume that there are two conda environments configured for either DMMN-ovary or CLAM):

cd <fullPath_to_DMMN-ovary_repo>
conda activate <clam_conda_env_name>
python slidereader_coords_dmmn2clam.py --source <path_to_slideList>.txt --out_path <coord_file_name>.csv
python inference_dmmn2clam.py --coord_path <coord_file_name>.csv --source <source_data> --out_path <seg_folder_name>
conda deactivate

cd <fullPath_to_CLAM_repo>
conda activate <dmmn_conda_env_name>
python create_patches_fp_dmmn2clam.py --source <source_data> --dmmn_seg <seg_folder_name>  --save_dir <output_dir> --patch_size 256 --seg --patch --stitch
conda deactivate

Issues

• I launched the segmentation from PyCharm. The following error was thrown which should be linked to a memory issue:

  Process finished with exit code 137 (interrupted by signal 9: SIGKILL)
  

• A couple of slide segmentations seem to have been stopped abruptly.

Results

Check here for some results.