search

mahmoodlab / CLAM

Data-efficient and weakly supervised computational pathology on whole slide images - Nature Biomedical Engineering
http://clam.mahmoodlab.org
GNU General Public License v3.0
1.07k stars 350 forks source link

Many Nan values in the extracted feature #225

Closed folkaholic closed 5 months ago

folkaholic commented 8 months ago

Hello there, thanks for sharing this useful tool. I have encounter a problem, I downloaded, patched, and extracted the feature of the TCGA dataset as told by the instruction. I checked the mask files which are normal, but I found there are so many Nan and inf values in the pt_files. Does anyone has similar problems? Thank you.

fedshyvana commented 6 months ago

do you have example data for reproducing this issue?

eternalld commented 6 months ago

I also encountered this problem, is there a solution now?

eternalld commented 6 months ago

do you have example data for reproducing this issue?

We use liver cancer pathological images from the TCGA database, for example, the file name is "TCGA-2Y-A9HB-01Z-00-DX1.ECF68E86-4D45-4545-A1D2-184D5ED38C3C" ， Thank You

eternalld commented 6 months ago

Sometimes I encounter an error when executing feature extract_fp.py: "openslide.lowlevel.OpenSlideError: Corrupt JPEG data: premature end of data segment", but when I use Qupath to open the file I find that the svs image can be opened normally. Have you ever encountered this situation? Thank you！！

fedshyvana commented 6 months ago

Regarding NaNs, I think i saw another post saying they changed the torch version and it was fixed. I recently bumped all the package dependencies to be much more up to date, could you give it a try to see if it solves your issue?

eternalld commented 6 months ago

okok，I'll try it.

---Original--- From: "Max @.> Date: Tue, Apr 9, 2024 12:00 PM To: @.>; Cc: @.**@.>; Subject: Re: [mahmoodlab/CLAM] Many Nan values in the extracted feature (Issue#225)

Regarding NaNs, I think i saw another post saying they changed the torch version and it was fixed. I recently bumped all the package dependencies to be much more up to date, could you give it a try to see if it solves your issue?

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>

fedshyvana commented 5 months ago

please let me know / re-open if you experience further issues on this, thanks