Pipeline for Extracting Features from Foundation Models

Environment Building

conda env create -f environment.yml

Step 1: Create a CSV file containing the paths of whole slide images (WSIs).

Run wsi_sheet/wsi_sheet.sh and modify the code if needed.

Step 2: Compute the coordinates of valid tiles.

Run compute_coords/build_run_scripts.sh for all sheets you created in Step 1.

• If you want, you can easily combine Step 1 & 2.

Step 3: Extract features using foundation models

In the feature_extraction folder, you can modify your own feature extraction code and the shell script to submit all extraction jobs at once.

In Step 2 and 3, we need a text file to list all cancers(jobs) name. The intuition is that we will iterate the list to build jobs and submit them. (Refer to tcga_all_cancer.txt file in folder)

I provide a simple requirements file, you can install some dependencies by pip3 install -r requirements.txt.

If you have any issues, feel free to reach out to me via email or Slack @Fang-Yi Su

Contact: fang-yi_su@hms.harvard.edu