Import the experimental virtual environment in conda

conda env -f mran.yaml 

then enter the environment:

conda activate mran 

Sample dataset download: Link 1 Link 2

After downloading, extract example.zip to the MRAN/WSI/ directory, then preprocess the original image:

cd pre python run_preprocess.py 

The storage format of the original data set can refer to the sample data set in MRAN/WSI/example :

MRAN/WSI/dataset_name/*/slide-idx1.svs

...

MRAN/WSI/dataset_name/*/slide-idxn.svs

Because the example dataset comes from TCGA, the first 12 bits of the file name slide_idxi.svs of each image are its case id.

cd pre python pro_csv.py 

The format of the label file of the original dataset can refer to MRAN/csv/example/sheet/total.csv:

python main.py 

main.yaml is used to set parameters.

python test.py 

The predict.csv of the directory "output_dir" records the prediction results on the test set.

cd interpretability python top_bag_top_patch.py 

"output_dir"/tb_tp/*/predict.csv records the prediction results on the test set.