MaskedPatchWSIDataset on unequal sample sizes

[a-parida12]

I am trying to some experiments with the Chamelyon Dataset using the MaskedPatchWSIDataset code below-

dir_path:str="/Datasets/CHAMELYON_IMG"
extn:str=".tif"
        
search_parameter = os.path.join(dir_path, '**/*'+extn)
train_data_list = [{"image": image_path} for image_path in glob.glob(search_parameter, recursive=True) ]
# load datasets and set the transforms
preprocess_cpu_train = Compose(
            [
                ScaleIntensityRangeD(keys="image", a_min=0.0, a_max=255.0, b_min=0, b_max=1.0),
                ToTensorD(keys="image"),
            ]
        )
        
dataset = MaskedPatchWSIDataset(
        data=train_data_list,
        patch_size=patch_size,
        transform=preprocess_cpu_train,
        reader="cuCIM",
        )
dataloader = DataLoader(dataset, batch_size=128, num_workers=2)

When I sample from the dataloader I am faced with the error below

  File "/home/abhijeet/Code/NST-Fed/.nstfed/lib/python3.11/site-packages/torch/utils/data/_utils/worker.py", line 308, in _worker_loop
    data = fetcher.fetch(index)
           ^^^^^^^^^^^^^^^^^^^^
  File "/home/abhijeet/Code/NST-Fed/.nstfed/lib/python3.11/site-packages/torch/utils/data/_utils/fetch.py", line 51, in fetch
    data = [self.dataset[idx] for idx in possibly_batched_index]
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/abhijeet/Code/NST-Fed/.nstfed/lib/python3.11/site-packages/torch/utils/data/_utils/fetch.py", line 51, in <listcomp>
    data = [self.dataset[idx] for idx in possibly_batched_index]
            ~~~~~~~~~~~~^^^^^
  File "/home/abhijeet/Code/NST-Fed/.nstfed/lib/python3.11/site-packages/torch/utils/data/dataset.py", line 298, in __getitem__
    return self.dataset[self.indices[idx]]
           ~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^
  File "/home/abhijeet/Code/NST-Fed/.nstfed/lib/python3.11/site-packages/monai/data/dataset.py", line 110, in __getitem__
    return self._transform(index)
           ^^^^^^^^^^^^^^^^^^^^^^
  File "/home/abhijeet/Code/NST-Fed/.nstfed/lib/python3.11/site-packages/monai/data/wsi_datasets.py", line 155, in _transform
    image, metadata = self._get_data(sample)
                      ^^^^^^^^^^^^^^^^^^^^^^
  File "/home/abhijeet/Code/NST-Fed/.nstfed/lib/python3.11/site-packages/monai/data/wsi_datasets.py", line 148, in _get_data
    return self.wsi_reader.get_data(wsi=wsi_obj, location=location, size=size, level=level)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/abhijeet/Code/NST-Fed/.nstfed/lib/python3.11/site-packages/monai/data/wsi_reader.py", line 256, in get_data
    raise ValueError(f"Location is outside of the image: location={location}, image size={wsi_size}")
ValueError: Location is outside of the image: location=[210607  36921], image size=(210579, 94968)

Do all the images have to be of the same size to use MaskedPatchWSIDataset?

[KumoLiu]

Hi @a-parida12, I don't think MaskedPatchWSIDataset must ensure all the images have the same size. May I know your MONAI version? Or maybe share one piece of data then I could help take a deep look at it, thanks!

[a-parida12]

I am using the weekly release version.
monai-weekly == 1.2.dev2313
commit-id : 795bf61adb4c25a03a355fa9c4552473eb763939

The data is too big to share here. You can download it from https://camelyon17.grand-challenge.org/Data/
Also, this problem has an issue #6360

[bhashemian]

Thanks @a-parida12 for your question. As we investigated in #6360, the error was not reproducible, and MaskedPatchWSIDataset works fine on input images with various sizes.