mahmoodlab
Problem when reproducing result
Hi, thanks for an amazing work. I am trying to reproduce the results reported in the paper and I am having this problem. So in the paper, it is said that only primary diagnosis WSIs are collected (I assume they are the one with DX1 in the slide_ids). So I downloaded and pre-processed these ones with TransPath.
However, when I try to run the script survpath.sh, I got this error
which means the code is loading the files not mentioned in the paper (this one containing DX2). I realized the problem is due to the fact that in this function, we are loading all the slides relating to a patient (which mean we are including the files not in primary diagnosis). So I wonder why is this the case? I found a work-around by doing this: temp_label_data = temp_label_data.groupby('case_id').first() (so now we only load 1 slide per patient). I am not sure if that work-around is correct because I am getting different results from what reported in the paper while running on BLCA.
Also, why encoding_dim is 1024 in survpath.sh? shouldn't it be 768?
