PC-TMB

Pytorch implementation of "Pan-cancer computational histopathology reveals tumor mutational burden status through weakly-supervised deep learning"

Step 1: Prepare WSI

Crop WSI into small patches and then extract low-dimensional farures. Refer to CLAM. We use customized self-supervised learning pretrained model rather than the routine ImageNet ptrained model. Weights and code of pretrained model will be releases upon publish.

Step 2: Construct graph using the embedding vectors

Use construct_graph.py

Step 3: Train MIL model.

Use train.py