Attention-Challenging Multiple Instance Learning for Whole Slide Image Classification

This is the Pytorch implementation of our "Attention-Challenging Multiple Instance Learning for Whole Slide Image Classification". This code is based on the CLAM.

Dataset Preparation

We provide a part of the extracted features to reimplement our results.

Extracted patch features using SSL ViT-S/16 on Camelyon16: https://drive.google.com/file/d/1sUxWoYlWooSq2M4fmgjADnDknzS5Fgaj/view?usp=drive_link

Extracted patch features using ImageNet supervised ResNet18 on Bracs: https://drive.google.com/file/d/1VaaaZF_anH46wQG2NxRAt4ENBQuNoXAq/view?usp=drive_link

Extracted patch features using SSL ViT-S/16 on Bracs: https://drive.google.com/file/d/14nYcr7SahYI0xSNVTXnp7u2ju3FSdrOU/view?usp=drive_link

For your own dataset, you can modify and run Step1_create_patches_fp.py and Step2_feature_extract.py. More details about this file can refer CLAM. Note that we recommend extracting features using SSL pretrained method. Our code using the checkpoints provided by Benchmarking Self-Supervised Learning on Diverse Pathology Datasets

Training

For the ABMIL (baseline), you should set n_token=1 n_masked_patch=0 mask_drop=0

CUDA_VISIBLE_DEVICES=2 python Step3_WSI_classification.py --seed 4 --wandb_mode online --arch ga --n_token 1 --n_masked_patch 0 --mask_drop 0 --config config/bracs_natural_supervised_config.yml

For our ACMIL, you should set n_token=5 n_masked_patch=10 mask_drop=0.6

CUDA_VISIBLE_DEVICES=2 python Step3_WSI_classification.py --seed 4 --wandb_mode online --arch ga --n_token 5 --n_masked_patch 10 --mask_drop 0.6 --config config/bracs_natural_supervised_config.yml