Class PydicomReader

monai/data/image_reader.py:393–997

Load medical images based on Pydicom library. All the supported image formats can be found at: https://dicom.nema.org/medical/dicom/current/output/chtml/part10/chapter_7.html PydicomReader is also able to load segmentations, if a dicom file contains tag: `SegmentSequence`, the read

@require_pkg(pkg_name="pydicom")
class PydicomReader(ImageReader):
    """
    Load medical images based on Pydicom library.
    All the supported image formats can be found at:
    https://dicom.nema.org/medical/dicom/current/output/chtml/part10/chapter_7.html

    PydicomReader is also able to load segmentations, if a dicom file contains tag: `SegmentSequence`, the reader
    will consider it as segmentation data, and to load it successfully, `PerFrameFunctionalGroupsSequence` is required
    for dicom file, and for each frame of dicom file, `SegmentIdentificationSequence` is required.
    This method refers to the Highdicom library.

    This class refers to:
    https://nipy.org/nibabel/dicom/dicom_orientation.html#dicom-affine-formula
    https://github.com/pydicom/contrib-pydicom/blob/master/input-output/pydicom_series.py
    https://highdicom.readthedocs.io/en/latest/usage.html#parsing-segmentation-seg-images

    Args:
        channel_dim: the channel dimension of the input image, default is None.
            This is used to set original_channel_dim in the metadata, EnsureChannelFirstD reads this field.
            If None, `original_channel_dim` will be either `no_channel` or `-1`.
        affine_lps_to_ras: whether to convert the affine matrix from "LPS" to "RAS". Defaults to ``True``.
            Set to ``True`` to be consistent with ``NibabelReader``,
            otherwise the affine matrix remains in the Dicom convention.
        swap_ij: whether to swap the first two spatial axes. Default to ``True``, so that the outputs
            are consistent with the other readers.
        prune_metadata: whether to prune the saved information in metadata. This argument is used for
            `get_data` function. If True, only items that are related to the affine matrix will be saved.
            Default to ``True``.
        label_dict: label of the dicom data. If provided, it will be used when loading segmentation data.
            Keys of the dict are the classes, and values are the corresponding class number. For example:
            for TCIA collection "C4KC-KiTS", it can be: {"Kidney": 0, "Renal Tumor": 1}.
        fname_regex: a regular expression to match the file names when the input is a folder.
            If provided, only the matched files will be included. For example, to include the file name
            "image_0001.dcm", the regular expression could be `".*image_(\\d+).dcm"`. Default to `""`.
            Set it to `None` to use `pydicom.misc.is_dicom` to match valid files.
        to_gpu: If True, load the image into GPU memory using CuPy and Kvikio. This can accelerate data loading.
            Default is False. CuPy and Kvikio are required for this option.
            In practical use, it's recommended to add a warm up call before the actual loading.
            A related tutorial will be prepared in the future, and the document will be updated accordingly.
        kwargs: additional args for `pydicom.dcmread` API. more details about available args:
            https://pydicom.github.io/pydicom/stable/reference/generated/pydicom.filereader.dcmread.html
            If the `get_data` function will be called
            (for example, when using this reader with `monai.transforms.LoadImage`), please ensure that the argument
            `stop_before_pixels` is `True`, and `specific_tags` covers all necessary tags, such as `PixelSpacing`,
            `ImagePositionPatient`, `ImageOrientationPatient` and all `pixel_array` related tags.
    """

    def __init__(
        self,
        channel_dim: str | int | None = None,
        affine_lps_to_ras: bool = True,
        swap_ij: bool = True,
        prune_metadata: bool = True,
        label_dict: dict | None = None,
        fname_regex: str = "",
        to_gpu: bool = False,
        **kwargs,
    ):