PANDA dataset noisy WSIs and dataset splits

[undercutspiky]

In the paper, you cite Kaggle and a paper for removing WSIs with noisy labels. I couldn't find a list of files with noisy labels that can be safely removed. Would it be possible for you to share the list of files you used?
Also, it'd be great if you could also share the exact train, val, and test splits used for PANDA so that I can just cite your work for using the specific WSIs and the split.

Thank you.

[HelloWorldLTY]

Same question here, I also wonder if it is possible to access datasets and paired labels used in this figure:

Thanks a lot!!!

[undercutspiky]

I was pointed to the splits used in their PANTHER paper by the corresponding author: https://github.com/mahmoodlab/PANTHER/tree/main/src/splits/classification/panda_wholesight

However, there are 2 problems:

1. The splits seem to be 60:20:20 instead of 80:10:10 (the PANTHER paper says that they are 80:10:10).

2. Minor: those splits have 9552 WSIs instead of 9555. PANTHER paper doesn't say anything about the total number of slides but this paper says that the total number of WSIs are supposed to be 9555.