Correct Way to Perform Preprocessing

[yuvfried]

Thanks for your great work.
Should I perform preprocessing on my own data before using the randstainna transform?
Or could I use your attached yaml as general configuration to H&E patches?
If the first option is true, should I calculate the stats on my whole patches dataset?

[yiqings]

@yuvfried
Thanks for your nice question.

To achieve better performance, we would recommend using a self-preprocessed yaml, and you could surely use the attached yaml as an initial attempt (but might not be the optimal one if turned into a new set).

To avoid test data leakage, the stats should only be calculated on your training patches dataset.