search

DIAGNijmegen / pathology-hooknet-tls

MIT License
6 stars 0 forks source link

Available spacing and downsampling factors of input TCGA files #3

Open yann-rdgz opened 4 months ago

yann-rdgz commented 4 months ago

Hello ! I tried to predict on TCGA cohorts I have this warning :

/workspace/env/hooknet/lib/python3.9/site-packages/wholeslidedata/image/wholeslideimage.py:78: UserWarning: spacing 0.5 outside margin (0.3%) for [0.2277, 0.9108, 3.6432, 14.5728], returning closest spacing: 0.2277

Then algorithm won't work properly, using mpp ~0.25 instead of 0.5.

I tested the prediction on the same slide as provided in the example, but using the original .svs file from TCGA-LUSC :

Using original .svs slide, I got the error:

  File "/workspace/pathology-hooknet-tls/hooknettls/__main__.py", line 8, in <module>
    objects = build_config(config_reader.read()["default"])
  File "/workspace/env/hooknet/lib/python3.9/site-packages/dicfg/factory.py", line 124, in build_config
    return _ObjectFactory(deepcopy(config)).build_config()
  File "/workspace/env/hooknet/lib/python3.9/site-packages/dicfg/factory.py", line 26, in build_config
    return self._build(self._configuration)
  File "/workspace/env/hooknet/lib/python3.9/functools.py", line 938, in _method
    return method.__get__(obj, cls)(*args, **kwargs)
  File "/workspace/env/hooknet/lib/python3.9/site-packages/dicfg/factory.py", line 38, in _build_dict
    config[key] = self._build_object(value)
  File "/workspace/env/hooknet/lib/python3.9/site-packages/dicfg/factory.py", line 67, in _build_object
    return attribute(*args, **kwargs)
  File "/workspace/env/hooknet/lib/python3.9/site-packages/wholeslidedata/iterators/patchiterator.py", line 46, in create_patch_iterator
    commander = commander_class(
  File "/workspace/env/hooknet/lib/python3.9/site-packages/wholeslidedata/buffer/patchcommander.py", line 41, in __init__
    wsi = WholeSlideImage(image_path, backend=backend)
  File "/workspace/env/hooknet/lib/python3.9/site-packages/wholeslidedata/image/wholeslideimage.py", line 35, in __init__
    self._backend = get_backend(backend)(path=self._path)
  File "/workspace/env/hooknet/lib/python3.9/site-packages/wholeslidedata/interoperability/asap/backend.py", line 15, in __init__
    raise ValueError(f"cant open image {path}")
ValueError: cant open image /opt/storage/nfs/raw/TCGA_BLCA/histology/raw/parafine/d09680b0-88fa-4814-99a7-c034d7fc43d5/TCGA-2F-A9KO-01Z-00-DX1.195576CF-B739-4BD9-B15B-4A70AE287D3E.svs

So changed the backend: asap to openslide to be able to open the file and I get the warning about spacing above!

I looked at the metadata of your .tif vs original one and they seem to have different downsampling factors wit level 0 with at a different mpp (spacing)

So my question was, to be able to predict did you transform all TCGA (BLCA, KIRC, LUSC) input .svs slides to have different downsampling factors inside .tif to have mpp=0.5 available ?

Thanks a lot for your help !! 🙏

martvanrijthoven commented 4 months ago

Dear Yann,

Yes, indeed, we resampled all the slides such that ~0.5um/px and ~2.0um/px become available, which are the mpps the algorithm requires for accurate predictions. Unfortunately , I haven't yet developed an on-the-fly resampling feature. However, I am actively working on this and hope to have it available shortly. In the interim, it is necessary to have these specific mpps available within the slides. I recommend using tools like pyvips or ASAP for converting to TIFF format, which ensures the slides are resampled and includes the required mpps. Please let me know if I can be of any assistance here.

Best wishes, Mart

yann-rdgz commented 4 months ago

Thank you for your precisions it helps a lot ! thank you for the support !