Question about "false_graph_filtering" function

[JaesikKim]

Hi. Thanks for your very interesting work!

I'm trying to replicate/reimplement your work on other cancer dataset.
But, I cannot get the underlying idea of this function

This is applied after generation of superpatch network, but without running this, it looks already generating good superpatch graph..
Can you please explain what and why applying this function?

And one more relevant question is about threshold.
The spatial threshold for edge is 5.5, and the theshold for false_graph_filtering is 4.3 in the code.
Can you explain why you set these values?
Accoding to the paper, the superpatch graph is defined within length of 5 tiles. How is this related to the thresholds you used?

(멋진연구 감사합니다. Feel free to answer in Korean if you prefer.. I don't mind at all!)
Jaesik

[liuxiaoping2020]

I have faced similar question:

When I run "supernode_generation.py" for the first step, it seems that it only run to

model_ft = model_ft.to(device)
model_ft.eval()
with tqdm(total=len(final_files)) as pbar_tot:
for image in final_files:
supernode_generation(image, model_ft, device, Argument, save_dir)
pbar_tot.update()

and generated 2 subdirectories, namely, "oringal", and "superpatch". In the "oringal" directory, it generated 3 kinds of files, examples of which are"TCGA-2F-A9KT-01Z-00-DX1_feature_list.csv", "TCGA-2F-A9KT-01Z-00-DX1_node_dict.pkl", "TCGA-2F-A9KT-01Z-00-DX1_node_location_list.csv", respectively. In the "superpatch" directory, it generated two kinds of files, examples of which are "TCGA-2F-A9KP-01Z-00-DX1_0.75_graph_torch.pt" and "TCGA-2F-A9KT-01Z-00-DX1_0.75.csv" respectively.

However, as you indicated, the output file should be: Compressed network as ".pt", Node position information in "_node_location_list.csv", Superpatch aggregated dictionary in "_artifact_sophis_final.csv". It seems that the final line (false_graph_filtering(4.3)) of "supernode_generation.py" did not work. I have checked the defination of "false_graph_filtering", and I found there is a "root_dir" specified through "root_dir = './Sample_data_for_demo/Graph_test/'" , however, I could not found a directory named "Graph_test" in the root directory. Thus, I guess the false_graph_filtering(4.3) might be work when we did not specified a true directory of "root_dir".

How to change/specify the "root_dir"? or can you help me address the problem?

Best regards

[liuxiaoping2020]

@JaesikKim can you run the funciton "false_graph_filtering", it seems that the "root_dir" in the function should be specified rather than the default value gived by the funciton. and can you tell me how you specified the root_dir to make the function work? thanks

[JaesikKim]

@liuxiaoping2020 In my case, I changed a few lines in false_graph_filtering to run just for a single file, and it ran well. You can take a look at my modified code if you want.

But, I still cannot get the idea behind this function. Do you have any idea?

false_graph_filtering.py.zip