how to get ./Downstream/Tumor_origin/src/feature/tcga/TCGA-LN-A8I1-01Z-00-DX1.F2C4FBC3-1FFA-45E9-9483-C3F1B2B7EF2D.pt

[27yw]

Hi! very wonderful work.
Here comes some question
how to get ./Downstream/Tumor_origin/src/feature/tcga/TCGA-LN-A8I1-01Z-00-DX1.F2C4FBC3-1FFA-45E9-9483-C3F1B2B7EF2D.pt

when I try to run this code
python3 Get_CHIEF_WSI_level_feature_batch.py

and I got this error
Original Traceback (most recent call last): File "/data_train/code/biz/yw/miniconda3/envs/chief2/lib/python3.9/site-packages/torch/utils/data/_utils/worker.py", line 351, in _worker_loop data = fetcher.fetch(index) # type: ignore[possibly-undefined] File "/data_train/code/biz/yw/miniconda3/envs/chief2/lib/python3.9/site-packages/torch/utils/data/_utils/fetch.py", line 54, in fetch data = self.dataset[possibly_batched_index] File "/data_train/code/biz/yw/CHIEF/datasets/BagDataset.py", line 25, in __getitem__ features = torch.load(full_path, map_location=torch.device('cpu')) File "/data_train/code/biz/yw/miniconda3/envs/chief2/lib/python3.9/site-packages/torch/serialization.py", line 1319, in load with _open_file_like(f, "rb") as opened_file: File "/data_train/code/biz/yw/miniconda3/envs/chief2/lib/python3.9/site-packages/torch/serialization.py", line 659, in _open_file_like return _open_file(name_or_buffer, mode) File "/data_train/code/biz/yw/miniconda3/envs/chief2/lib/python3.9/site-packages/torch/serialization.py", line 640, in __init__ super().__init__(open(name, mode)) FileNotFoundError: [Errno 2] No such file or directory: './Downstream/Tumor_origin/src/feature/tcga/TCGA-B6-A0WV-01Z-00-DX1.A8B9E114-A8CF-4389-B47C-2E1B842F7FF9.pt'
how to fix it?

[27yw]

btw there is no feature/tcga folder under Downstream/Tumor_origin/src/

[Dadatata-JZ]

@27yw thanks.

It is just intended to show how to load a feature file. It is really nothing technical.
You should (1) tile your WSIs, (2) extract tile-level features (3) stack the features as a matrix in a file (e.g., pickle). "

You could refer to ctranspath repo to catch up with some prior.
https://github.com/Xiyue-Wang/TransPath

[27yw]

HI thanks for your answering
so you use this repo to get the features and save into a .pt file?
https://github.com/Xiyue-Wang/TransPath

[ivicts]

@Dadatata-JZ Hi for survial, inference.ipynb, may I ask if the feature file is tile-level features (Get_CHIEF_patch_feature.py) or WSI level features (Get_CHIEF_WSI_level_feature.py)?

[Dadatata-JZ]

@ivicts You would need both to obtain the tile-level and then slide-level features.

[Dadatata-JZ]

HI thanks for your answering so you use this repo to get the features and save into a .pt file? https://github.com/Xiyue-Wang/TransPath

@27yw Yes, you can.

[ivicts]

@Dadatata-JZ Thanks, I just want to run the inference.ipynb, for the survival problem for my custom data. For this problem, what are the input features? Is it:

1. Tile-level features (Get_CHIEF_patch_feature.py)
2. WSI level features (Get_CHIEF_WSI_level_feature.py)
3. (a) Tile my WSIs
   (b) Extract tile-level features (Get_CHIEF_patch_feature.py)
   (c) Stack the features as a matrix in a file (e.g., pickle, concatenate? what dimension?).

Which one should I do? I just want to run the survival analysis for my custom data.

[ivicts]

@Dadatata-JZ are there any updates on this?

[Dadatata-JZ]

@ivicts

4.
(a) Tile my WSIs
(b) Extract tile-level features (Get_CHIEF_patch_feature.py or any other tile-level feature extractors)
(c) concat tile-level features then extract slide-level features
(d) tune it for patient stratification (i.e., risk score/group up to your applications)
(e) get the corresponding survival information, then compute KM curves and check c-index etc.

Also, if you are not familiar with survival analysis using deep learning/CPath, some previous works may be inspiring. for instance,
https://pubmed.ncbi.nlm.nih.gov/38123254/

Feel free to shoot me an email and loop me into your real applications if prompt responses are expected. working on responding grant calls and other manuscripts. I don't get on GITHUB as often as before.