clemsgrs / hs2p

Histopathology Slides Preprocessing Pipeline
30 stars 7 forks source link

Error in Tiling Some TCGA NSCLC Slides #17

Open bryanwong17 opened 1 month ago

bryanwong17 commented 1 month ago

Hi @clemsgrs , I encountered errors while extracting patches with hs2p on TCGA NSCLC dataset, making it impossible to tile some slides. The errors are detailed in the attached process_list.csv. Could you please check it and advise on how to resolve thse issues to tile all slides? Thanks in advance!

process_list.csv

clemsgrs commented 1 month ago

Hi, it seems the error stems from openslide. You could try to use backend: "pyvips" in your config file to see if switching the image opening backend solves the error.

bryanwong17 commented 1 month ago

I tried changing the backend to 'pyvips', but it did not work. I will re-download the slides and try using the 'openslide' backend again. My assumption is that some of the slides might be corrupted. I will let you know later.

bryanwong17 commented 1 month ago

Hi @clemsgrs , I have tried redownloading the files, but it is still not working. My guess is that for some slides, the spacing = float(self.properties["openslide.mpp-x"]) property cannot be retrieved, which results in an error. Do you have any suggestions on how I can still extract the patches despite the missing spacing information?

clemsgrs commented 1 month ago

Hi, when the spacing is missing / off, you can manually specify the spacing at level 0 in the input csv. Simply add a "spacing" column with the spacing at level 0 for each slide. Let me know if it solves your error.

bryanwong17 commented 1 month ago

Hi @clemsgrs , I tried adding a 'spacing' column as shown below, but I still could not extract the patches.

slide_id,slide_path,spacing
TCGA-05-4390-01Z-00-DX1,/vast/WSI_datasets/tcga_nsclc_new/WSI/LUAD/TCGA-05-4390-01Z-00-DX1.svs,0.25
TCGA-05-5428-01Z-00-DX1,/vast/WSI_datasets/tcga_nsclc_new/WSI/LUAD/TCGA-05-5428-01Z-00-DX1.svs,0.25
TCGA-05-5429-01Z-00-DX1,/vast/WSI_datasets/tcga_nsclc_new/WSI/LUAD/TCGA-05-5429-01Z-00-DX1.svs,0.25