KatherLab / STAMP

Solid Tumor Associative Modeling in Pathology
MIT License
47 stars 23 forks source link

Unsupported TIFF compression: 7 #27

Closed suminwei closed 5 months ago

suminwei commented 5 months ago

Hi, I'm running stamp preprocess and I'm using .svs files downloaded from TCGA as input. However, I keep on getting the error

Failed loading slide, continuing... Error: Unsupported TIFF compression: 7

From reading the protocol online, STAMP supports SVS file formats. I also tried .tiff and .jpeg, but nothing seems to work. I'd appreciate your help! Thank you, Lisa

suminwei commented 5 months ago

For setting up STAMP, I generally followed the instructions on Github to set up the STAMP conda environment. I work in a HPC computing environment, linux system, CPU/GPU available. For the config.yaml file, I only modified the Preprocessing section so far from the original template config file. I changed it to this (I took out the full paths but they are all absolute paths in the actual file):


preprocessing:
  output_dir: /path/to/STAMP/out_dir # Path to save features to
  wsi_dir: /path/to/STAMP/input # Path of where the whole-slide images are.
  cache_dir: /path/to/STAMP/cache_dir # Directory to store intermediate slide JPGs
  microns: 256 # Edge length in microns for each patch (default is 256, with pixel size 224, 256/224 = ~1.14MPP = ~9x magnification)
  norm: false # Perform Macenko normalisation
  feat_extractor: ctp # Use ctp for CTransPath (default) or uni for UNI (requires prior authentication)
  del_slide: false # Remove the original slide after processing
  cache: true # Save intermediate images (slide, background rejected, normalized)
  only_feature_extraction: false # Only perform feature extraction (intermediate images (background rejected, [normalized]) have to exist)
  cores: 8 # CPU cores to use
  device: cuda:0 # device to run feature extraction on (cpu, cuda, cuda:0, etc.)
Avic3nna commented 5 months ago

Solved - Above error was produce on HPC and their local OpenSlide installation was likely not the right version. Then, used the container (built from container.def) and problems were resolved.