Datasets

Dataset classes are required by the WSIInference class to streamline the process of patch extraction based on user-defined parameters. Two WSI dataset classes, available in dplabtools, integrate with patch location classes introduced in Patch locations/sampling.

Note

Since WSI dataset classes are descendants of PyTorch Dataset class, correctly installed PyTorch is required to import them in Python code.

WSI dataset class

class dplabtools.slides.processing.WSIDataset(...)

Class for patch extraction during inference.

Basic usage

Assuming that patches object represents one of the classes introduced in Patch locations/sampling:

from dplabtools.slides.processing import WSIDataset

dataset = WSIDataset(patches=patches)

See also

Full example of inference process using WSIDataset

Class details

WSIDataset has no class specific parameters.

See also

Parameters common in all WSI dataset classes

See also

Methods and properties common to all WSI dataset classes

WSI dataset class (MRP)

class dplabtools.slides.processing.WSIMultiResDataset(...)

Class for multi resolution patch extraction during inference.

See also

Multi resolution patches (MRP)

Basic usage

Assuming that patches object represents one of the classes introduced in Patch locations/sampling:

from dplabtools.slides.processing import WSIMultiResDataset

dataset = WSIMultiResDataset(patches=patches, levels_or_mpps=[0.5, 0.7. 0.9])

Class details

Parameters specific to WSIMultiResDataset:

class dplabtoolshiddenclass_35e0e776aad6467593b2bfe568e1db6e
Parameters:

levels_or_mpps (list of level_or_mpp values) – Numbers representing WSI levels or MPP values for multi resolution patches.

See also

Parameters common in all WSI dataset classes

See also

Methods and properties common to all WSI dataset classes

See also

level_or_mpp

Parameters common in all WSI dataset classes

class dplabtoolshiddenclass_84cbea39ef314baea74f83c892e9ccae
Parameters:

  • patches (object) – Object representing one of the patch location classes.

  • transform_fn (callable, optional) – A user-defined image transformation that will be called on each patch extracted via get_region, this transformation should run its own to-tensor conversion. If no transformation is provided, the image objects will be converted to tensors using to_tensor from the torchvision package.

  • resampling_mode (str, optional) – One of two supported down/up-sampling methods: wsi or tile.

  • extra_mpps (list of float, optional) – List of MPP values for wsi resampling mode.

  • zero_workers (bool, default=False) – Set to True, if WSIInference class will have num_workers set to 0.

  • save_patches_dir (str, optional) – Directory for saving the extracted patches, should only be used for troubleshooting inference problems. The type of saved image files can be changed using save_patches_image_type class attribute, the default image type is PNG.

Methods and properties common to all WSI dataset classes

Common methods and properties are derived from the base class.

class dplabtools.slides.processing.dataset.base.BaseDataset(...)

Base class for the WSI dataset classes.

property patches

Return the patches object used to build the dataset.

save_patches_image_type = 'png'

Class attribute for setting the image type when save_patches_dir is defined.