Confusions of implementation details in paper #2
Description
Hello, Dr. Yao, we have read your paper published on MIA. Your DeepAttnMISL model is a good method. We are also conducting related research and are trying to reproduce the DeepAttnMISL model on the NLST data set.We have some doubts about the method in the paper and would like to ask you for help.
Sampling of patches:
- Which layer of the wsi is sampled and how big is the size of the sampling patches
- Do the patches all come from the cancerous area, or are they randomly sampled from the slide?
- What method is used for sampling
- What is the number of sampling patches for each slide?
Determination of the survival time of each patient:
The NLST data set gives the following two variables：
candx_days: Days from randomization to first diagnosis of lung cancer
fup_days: Days from randomization to date last known alive
The calculation of survival time should use fup_days-candx_days, or use fup_days directly。