Class WSIReader

monai/data/wsi_reader.py:512–737

Read whole slide images and extract patches using different backend libraries Args: backend: the name of backend whole slide image reader library, the default is cuCIM. level: the whole slide image level at which the patches are extracted. mpp: the resolution in mic

class WSIReader(BaseWSIReader):
    """
    Read whole slide images and extract patches using different backend libraries

    Args:
        backend: the name of backend whole slide image reader library, the default is cuCIM.
        level: the whole slide image level at which the patches are extracted.
        mpp: the resolution in micron per pixel at which the patches are extracted.
        mpp_rtol: the acceptable relative tolerance for resolution in micro per pixel.
        mpp_atol: the acceptable absolute tolerance for resolution in micro per pixel.
        power: the objective power at which the patches are extracted.
        power_rtol: the acceptable relative tolerance for objective power.
        power_atol: the acceptable absolute tolerance for objective power.
        channel_dim: the desired dimension for color channel. Default to 0 (channel first).
        dtype: the data type of output image. Defaults to `np.uint8`.
        device: target device to put the extracted patch. Note that if device is "cuda"",
            the output will be converted to torch tenor and sent to the gpu even if the dtype is numpy.
        mode: the output image color mode, "RGB" or "RGBA". Defaults to "RGB".
        num_workers: number of workers for multi-thread image loading (cucim backend only).
        kwargs: additional arguments to be passed to the backend library

        Notes:
            Only one of resolution parameters, `level`, `mpp`, or `power`, should be provided.
            If such parameters are provided in `get_data` method, those will override the values provided here.
            If none of them are provided here or in `get_data`, `level=0` will be used.

    """

    supported_backends = ["cucim", "openslide", "tifffile"]

    def __init__(
        self,
        backend="cucim",
        level: int | None = None,
        mpp: float | tuple[float, float] | None = None,
        mpp_rtol: float = 0.05,
        mpp_atol: float = 0.0,
        power: int | None = None,
        power_rtol: float = 0.05,
        power_atol: float = 0.0,
        channel_dim: int = 0,
        dtype: DtypeLike | torch.dtype = np.uint8,
        device: torch.device | str | None = None,
        mode: str = "RGB",
        **kwargs,
    ):
        self.backend = backend.lower()
        self.reader: CuCIMWSIReader | OpenSlideWSIReader | TiffFileWSIReader
        if self.backend == "cucim":
            self.reader = CuCIMWSIReader(
                level=level,
                mpp=mpp,
                mpp_rtol=mpp_rtol,
                mpp_atol=mpp_atol,
                power=power,
                power_rtol=power_rtol,
                power_atol=power_atol,
                channel_dim=channel_dim,