pydicom.dataset.Dataset¶

class pydicom.dataset.Dataset(*args: Dataset | MutableMapping[BaseTag, DataElement | RawDataElement], **kwargs: Any)[source]¶

A DICOM dataset as a mutable mapping of DICOM Data Elements.

Examples

Add an element to the Dataset (for elements in the DICOM dictionary):

>>> ds = Dataset()
>>> ds.PatientName = "CITIZEN^Joan"
>>> ds.add_new(0x00100020, 'LO', '12345')
>>> ds[0x0010, 0x0030] = DataElement(0x00100030, 'DA', '20010101')

Add a sequence element to the Dataset

>>> ds.BeamSequence = [Dataset(), Dataset(), Dataset()]
>>> ds.BeamSequence[0].Manufacturer = "Linac, co."
>>> ds.BeamSequence[1].Manufacturer = "Linac and Sons, co."
>>> ds.BeamSequence[2].Manufacturer = "Linac and Daughters, co."

Add private elements to the Dataset

>>> block = ds.private_block(0x0041, 'My Creator', create=True)
>>> block.add_new(0x01, 'LO', '12345')

Updating and retrieving element values:

>>> ds.PatientName = "CITIZEN^Joan"
>>> ds.PatientName
'CITIZEN^Joan'
>>> ds.PatientName = "CITIZEN^John"
>>> ds.PatientName
'CITIZEN^John'

Retrieving an element’s value from a Sequence:

>>> ds.BeamSequence[0].Manufacturer
'Linac, co.'
>>> ds.BeamSequence[1].Manufacturer
'Linac and Sons, co.'

Accessing the DataElement items:

>>> elem = ds['PatientName']
>>> elem
(0010,0010) Patient's Name                      PN: 'CITIZEN^John'
>>> elem = ds[0x00100010]
>>> elem
(0010,0010) Patient's Name                      PN: 'CITIZEN^John'
>>> elem = ds.data_element('PatientName')
>>> elem
(0010,0010) Patient's Name                      PN: 'CITIZEN^John'

Accessing a private DataElement item:

>>> block = ds.private_block(0x0041, 'My Creator')
>>> elem = block[0x01]
>>> elem
(0041,1001) Private tag data                    LO: '12345'
>>> elem.value
'12345'

Alternatively:

>>> ds.get_private_item(0x0041, 0x01, 'My Creator').value
'12345'

Deleting an element from the Dataset

>>> del ds.PatientID
>>> del ds.BeamSequence[1].Manufacturer
>>> del ds.BeamSequence[2]

Deleting a private element from the Dataset

>>> block = ds.private_block(0x0041, 'My Creator')
>>> if 0x01 in block:
...     del block[0x01]

Determining if an element is present in the Dataset

>>> 'PatientName' in ds
True
>>> 'PatientID' in ds
False
>>> (0x0010, 0x0030) in ds
True
>>> 'Manufacturer' in ds.BeamSequence[0]
True

Iterating through the top level of a Dataset only (excluding Sequences):

>>> for elem in ds:
...    print(elem)
(0010,0010) Patient's Name                      PN: 'CITIZEN^John'

Iterating through the entire Dataset (including Sequences):

>>> for elem in ds.iterall():
...     print(elem)
(0010,0010) Patient's Name                      PN: 'CITIZEN^John'

Recursively iterate through a Dataset (including Sequences):

>>> def recurse(ds):
...     for elem in ds:
...         if elem.VR == 'SQ':
...             [recurse(item) for item in elem.value]
...         else:
...             # Do something useful with each DataElement

Converting the Dataset to and from JSON:

>>> ds = Dataset()
>>> ds.PatientName = "Some^Name"
>>> jsonmodel = ds.to_json()
>>> ds2 = Dataset()
>>> ds2.from_json(jsonmodel)
(0010,0010) Patient's Name                      PN: 'Some^Name'

default_element_format¶

The default formatting for string display.

Type:: str

default_sequence_element_format¶

The default formatting for string display of sequences.

Type:: str

indent_chars¶

For string display, the characters used to indent nested Sequences. Default is " ".

Type:: str

__init__(*args: Dataset | MutableMapping[BaseTag, DataElement | RawDataElement], **kwargs: Any) → None[source]¶: Create a new Dataset instance.

Methods

`__init__`(args, *kwargs)	Create a new `Dataset` instance.
`add`(data_element)	Add an element to the `Dataset`.
`add_new`(tag, VR, value)	Create a new element and add it to the `Dataset`.
`add_new_private`(private_creator, group, ...)	Create a new private element and add it to the `Dataset`.
`clear`()	Delete all the elements from the `Dataset`.
`compress`(transfer_syntax_uid[, arr, ...])	Compress uncompressed pixel data and update ds in-place with the resulting encapsulated codestream.
`convert_pixel_data`([handler_name])	Convert pixel data to a `numpy.ndarray` internally.
`copy`()	Return a shallow copy of the dataset.
`data_element`(name)	Return the element corresponding to the element keyword name.
`decode`()	Apply character set decoding to the elements in the `Dataset`.
`decompress`([handler_name, as_rgb, ...])	Perform an in-place decompression of a dataset with a compressed Transfer Syntax UID.
`dir`(*filters)	Return an alphabetical list of element keywords in the `Dataset`.
`elements`()	Yield the top-level elements of the `Dataset`.
`ensure_file_meta`()	Create an empty `Dataset.file_meta` if none exists.
`formatted_lines`([element_format, ...])	Iterate through the `Dataset` yielding formatted `str` for each element.
`from_json`(json_dataset[, bulk_data_uri_handler])	Return a `Dataset` from a DICOM JSON Model object.
`get`()	Simulate `dict.get()` to handle element tags and keywords.
`get_item`()	Return the raw data element if possible.
`get_private_item`(group, element_offset, ...)	Return the data element for the given private tag group.
`group_dataset`(group)	Return a `Dataset` containing only elements of a certain group.
`items`()	Return the `Dataset` items to simulate `dict.items()`.
`iterall`()	Iterate through the `Dataset`, yielding all the elements.
`keys`()	Return the `Dataset` keys to simulate `dict.keys()`.
`overlay_array`(group)	Return the Overlay Data in group as a `numpy.ndarray`.
`pixel_array_options`(*[, index, raw, ...])	Set the decoding and processing options used by the `pixel_array` property.
`pop`(key, *args)	Emulate `dict.pop()` with support for tags and keywords.
`popitem`()	Emulate `dict.popitem()`.
`private_block`(group, private_creator[, create])	Return the block for the given tag group and private_creator.
`private_creators`(group)	Return a list of private creator names in the given group.
`remove_private_tags`()	Remove all private elements from the `Dataset`.
`save_as`(filename, /[, ...])	Encode the current `Dataset` and write it to filename.
`set_original_encoding`(is_implicit_vr, ...[, ...])	Set the values for the original dataset encoding.
`set_pixel_data`(arr, ...[, generate_instance_uid])	Use an `ndarray` to set the Pixel Data and related Image Pixel module elements.
`setdefault`(key[, default])	Emulate `dict.setdefault()` with support for tags and keywords.
`to_json`([bulk_data_threshold, ...])	Return a JSON representation of the `Dataset`.
`to_json_dict`([bulk_data_threshold, ...])	Return a dictionary representation of the `Dataset` conforming to the DICOM JSON Model as described in the DICOM Standard, Part 18, Annex F.
`top`()	Return a `str` representation of the top level elements.
`trait_names`()	Return a `list` of valid names for auto-completion code.
`update`(d)	Extend `dict.update()` to handle DICOM tags and keywords.
`update_raw_element`(tag, *[, vr, value])	Modify the VR or value for the raw element with tag.
`values`()	Return the `Dataset` values to simulate `dict.values()`.
`walk`(callback[, recursive])	Iterate through the `Dataset's` elements and run callback on each.
`waveform_array`(index)	Return an `ndarray` for the multiplex group at index in the (5400,0100) Waveform Sequence.

Attributes

`default_element_format`
`default_sequence_element_format`
`indent_chars`
`is_decompressed`	Return `True` if the dataset uses an uncompressed (native) transfer syntax, `False` otherwise.
`is_implicit_VR`	Get/set the VR method used when encoding the dataset.
`is_little_endian`	Get/set the endianness used when encoding the dataset.
`is_original_encoding`	Return `True` if the encoding to be used for writing is set and is the same as that used to originally encode the `Dataset`.
`original_character_set`	Return the original character set encoding for a dataset decoded from a file or buffer.
`original_encoding`	Return the original encoding used for a dataset decoded from a file or buffer.
`pixel_array`	Return the pixel data as a `numpy.ndarray`.
`read_encoding`	Return the original character set encoding for a decoded dataset.
`read_implicit_vr`	Get the VR method used by the original encoding of the dataset.
`read_little_endian`	Get the endianness used by the original encoding of the dataset.

add(data_element: DataElement) → None[source]¶

Add an element to the Dataset.

Equivalent to ds[data_element.tag] = data_element

Parameters:: data_element (dataelem.DataElement) – The DataElement to add.

add_new(tag: int | str | tuple[int, int] | BaseTag, VR: str, value: Any) → None[source]¶

Create a new element and add it to the Dataset.

Parameters:

tag – The DICOM (group, element) tag in any form accepted by Tag() such as [0x0010, 0x0010], (0x10, 0x10), 0x00100010, etc.
VR (str) – The 2 character DICOM value representation (see DICOM Standard, Part 5, Section 6.2).
value –
The value of the data element. One of the following:
- a single string or number
- a list or tuple with all strings or all numbers
- a multi-value string with backslash separator
- for a sequence element, an empty list or list of Dataset

add_new_private(private_creator: str, group: int, element_offset: int, value: Any, vr: str | None = None) → None[source]¶

Create a new private element and add it to the Dataset.

Parameters:

private_creator (str) – The private creator string related to the new tag.
group (int) – The group ID (0x0009 - 0xFFFF) for the private tag. Must be an odd number.
element_offset (int) – The tag offset, e.g. the lower byte of the tag element of the private tag (0x00 - 0xFF). The higher byte is defined by the location of the private creator tag.
value (Any) –
The value of the data element. One of the following:
- a single string or number
- a list or tuple with all strings or all numbers
- a multi-value string with backslash separator
- for a sequence element, an empty list or list of Dataset
vr (str | None) – The two-letter DICOM value representation, or None. If set to None, it is taken from the private tag dictionary.

Raises:

ValueError – If group doesn’t belong to a private tag or private_creator is empty.
KeyError – If vr is None and the tag is not found in the private tag dictionary.

clear() → None[source]¶: Delete all the elements from the Dataset.

compress(transfer_syntax_uid: str, arr: ndarray | None = None, encoding_plugin: str = '', encapsulate_ext: bool = False, *, generate_instance_uid: None | bool = None, jls_error: int | None = None, j2k_cr: list[float] | None = None, j2k_psnr: list[float] | None = None, **kwargs: Any) → None[source]¶

Compress uncompressed pixel data and update ds in-place with the resulting encapsulated codestream.

Added in version 2.2.

The dataset ds must already have the following Image Pixel module elements present with correct values that correspond to the resulting compressed pixel data:

(0028,0002) Samples per Pixel
(0028,0004) Photometric Interpretation
(0028,0008) Number of Frames (if more than 1 frame will be present)
(0028,0010) Rows
(0028,0011) Columns
(0028,0100) Bits Allocated
(0028,0101) Bits Stored
(0028,0103) Pixel Representation

If Samples per Pixel is greater than 1 then the following element is also required:

(0028,0006) Planar Configuration

This method will add the file meta dataset if none is present and add or modify the following elements:

(0002,0010) Transfer Syntax UID
(7FE0,0010) Pixel Data

If the compressed pixel data is too large for encapsulation using a basic offset table then an extended offset table will also be used, in which case the following elements will also be added:

(7FE0,0001) Extended Offset Table
(7FE0,0002) Extended Offset Table Lengths

If performing lossy compression then a new (0008,0018) SOP Instance UID value will be generated unless generate_instance_uid is False.

Supported Transfer Syntax UIDs

UID		Plugins	Encoding Guide
Name	Value	Plugins	Encoding Guide
JPEG-LS Lossless	1.2.840.10008.1.2.4.80	pyjpegls	JPEG-LS
JPEG-LS Near Lossless	1.2.840.10008.1.2.4.81	pyjpegls	JPEG-LS
JPEG 2000 Lossless	1.2.840.10008.1.2.4.90	pylibjpeg	JPEG 2000
JPEG 2000	1.2.840.10008.1.2.4.91	pylibjpeg	JPEG 2000
RLE Lossless	1.2.840.10008.1.2.5	pydicom, pylibjpeg, gdcm	RLE Lossless
Deflated Image Frame Compression	1.2.840.10008.1.2.8.1	pydicom	Deflated Image

Changed in version 3.0: Added the jls_error, j2k_cr, j2k_psnr and generate_instance_uid keyword parameters.

Changed in version 3.1: Added support for Deflated Image Frame Compression

Examples

Compress the existing uncompressed Pixel Data in place:

>>> from pydicom import examples
>>> from pydicom.uid import RLELossless
>>> ds = examples.ct
>>> ds.compress(RLELossless)
>>> ds.save_as("ct_rle_lossless.dcm")

Parameters:

transfer_syntax_uid (pydicom.uid.UID) – The UID of the transfer syntax to use when compressing the pixel data.
arr (numpy.ndarray, optional) – Compress the uncompressed pixel data in arr and use it to set the Pixel Data. If arr is not used then the existing Pixel Data in the dataset will be compressed instead. The shape, dtype and contents of the array should match the dataset.
encoding_plugin (str, optional) – Use the encoding_plugin to compress the pixel data. See the user guide for a list of plugins available for each UID and their dependencies. If not specified then all available plugins will be tried (default).
encapsulate_ext (bool, optional) – If True then force the addition of an extended offset table. If False (default) then an extended offset table will be added if needed for large amounts of compressed Pixel Data, otherwise just the basic offset table will be used.
generate_instance_uid (bool, optional) – If True then always generate a new (0008,0018) SOP Instance UID using generate_uid(), otherwise False to always keep the original. The default behavior is to only generate a new SOP Instance UID when performing lossy compression.
jls_error (int, optional) – JPEG-LS Near Lossless only. The allowed absolute compression error in the pixel values.
j2k_cr (list[float], optional) – JPEG 2000 only. A list of the compression ratios to use for each quality layer. There must be at least one quality layer and the minimum allowable compression ratio is 1. When using multiple quality layers they should be ordered in decreasing value from left to right. For example, to use 2 quality layers with 20x and 5x compression ratios then j2k_cr should be [20, 5]. Cannot be used with j2k_psnr.
j2k_psnr (list[float], optional) – JPEG 2000 only. A list of the peak signal-to-noise ratios (in dB) to use for each quality layer. There must be at least one quality layer and when using multiple quality layers they should be ordered in increasing value from left to right. For example, to use 2 quality layers with PSNR of 80 and 300 then j2k_psnr should be [80, 300]. Cannot be used with j2k_cr.
**kwargs – Optional keyword parameters for the encoding plugin may also be present. See the encoding plugins options for more information.

convert_pixel_data(handler_name: str = '') → None[source]¶

Convert pixel data to a numpy.ndarray internally.

Deprecated since version 3.0: This method will be removed in v4.0, use pixel_array_options() instead.

Parameters:

handler_name (str, optional) –

The name of the pixel handler or decoding plugin to use to decode the dataset’s pixel data. Support values are:

If using the pixel_data_handlers backend: 'gdcm', 'pillow', 'jpeg_ls', 'rle', 'numpy' and 'pylibjpeg'.
If using the pixels backend see the documentation for the decoder corresponding to the dataset’s Transfer Syntax UID.

If not used (the default) then all available handlers or plugins will be tried and the data from first successful one will be used.

Returns:

Converted pixel data is stored internally in the dataset, it can be accessed with the pixel_array property.

Return type:

None

Raises:

ValueError – If handler_name is not a valid handler name.
NotImplementedError – If the given handler or any handler, if none given, is unable to decompress pixel data with the current transfer syntax
RuntimeError – If the given handler, or the handler that has been selected if none given, is not available.

Notes

If the pixel data is in a compressed image format, the data is decompressed and any related data elements are changed accordingly.

copy() → Dataset[source]¶: Return a shallow copy of the dataset.

data_element(name: str) → DataElement | None[source]¶

Return the element corresponding to the element keyword name.

Parameters:: name (str) – A DICOM element keyword.
Returns:: For the given DICOM element keyword, return the corresponding DataElement if present, None otherwise.
Return type:: dataelem.DataElement or None

decode() → None[source]¶

Apply character set decoding to the elements in the Dataset.

See DICOM Standard, Part 5, Section 6.1.1.

decompress(handler_name: str = '', *, as_rgb: bool = True, generate_instance_uid: None | bool = None, decoding_plugin: str = '', **kwargs: Any) → None[source]¶

Perform an in-place decompression of a dataset with a compressed Transfer Syntax UID.

Warning

This function requires NumPy and may require the installation of additional packages to perform the actual pixel data decoding. See the pixel data decompression documentation for more information.

The dataset’s Transfer Syntax UID will be set to Explicit VR Little Endian.
The Pixel Data will be decompressed in its entirety and the Pixel Data element’s value updated with the decompressed data, padded to an even length.
The Pixel Data element’s VR will be set to OB if Bits Allocated <= 8, otherwise it will be set to OW.
The DataElement.is_undefined_length attribute for the Pixel Data element will be set to False.
Any image pixel module elements may be modified as required to match the uncompressed Pixel Data.
If the color space has been converted from YCbCr to RGB then a new (0008,0018) SOP Instance UID value will be generated unless generate_instance_uid is False.

Changed in version 3.0: Added the as_rgb and generate_instance_uid keyword parameters.

Deprecated since version 3.0: The handler_name parameter will be removed in v4.0, use decoding_plugin instead.

Parameters:

handler_name (str, optional) – Deprecated and will be removed in v4.0, use decoding_plugin instead.
as_rgb (bool, optional) – pixels backend only. If True (default) then convert pixel data with a YCbCr photometric interpretation such as "YBR_FULL_422" to RGB.
generate_instance_uid (bool, optional) – If True then always generate a new (0008,0018) SOP Instance UID, using generate_uid(), otherwise False to always keep the original. The default behavior is to only generate a new SOP Instance UID when the image data has had a YCbCr to RGB conversion applied.
decoding_plugin (str, optional) –
The name of the decoding plugin to use when decoding compressed pixel data. If no decoding_plugin is specified (default) then all available plugins will be tried and the result from the first successful one yielded.
- If using the pixels backend (default) then see the API documentation for the available plugins for each Transfer Syntax UID.
- If using the deprecated pixel_data_handlers backend supported plugins are: 'gdcm', 'pillow', 'jpeg_ls', 'rle', 'numpy' and 'pylibjpeg'.
kwargs (dict[str, Any], optional) – pixels backend only. Optional keyword parameters for the decoding plugin may also be present. See the decoding plugins options for more information.

dir(*filters: str) → list[str][source]¶

Return an alphabetical list of element keywords in the Dataset.

Intended mainly for use in interactive Python sessions. Only lists the element keywords in the current level of the Dataset (i.e. the contents of any sequence elements are ignored).

Parameters:: filters (str) – Zero or more string arguments to the function. Used for case-insensitive match to any part of the DICOM keyword.
Returns:: The matching element keywords in the dataset. If no filters are used then all element keywords are returned.
Return type:: list of str

elements() → Iterator[DataElement | RawDataElement][source]¶

Yield the top-level elements of the Dataset.

Examples

>>> ds = Dataset()
>>> for elem in ds.elements():
...     print(elem)

The elements are returned in the same way as in Dataset.__getitem__().

Yields:: dataelem.DataElement or dataelem.RawDataElement – The unconverted elements sorted by increasing tag order.

ensure_file_meta() → None[source]¶: Create an empty Dataset.file_meta if none exists.

formatted_lines(element_format: str = '%(tag)s %(name)-35.35s %(VR)s: %(repval)s', sequence_element_format: str = '%(tag)s %(name)-35.35s %(VR)s: %(repval)s', indent_format: str | None = None) → Iterator[str][source]¶

Iterate through the Dataset yielding formatted str for each element.

Parameters:

element_format (str) – The string format to use for non-sequence elements. Formatting uses the attributes of DataElement. Default is "%(tag)s %(name)-35.35s %(VR)s: %(repval)s".
sequence_element_format (str) – The string format to use for sequence elements. Formatting uses the attributes of DataElement. Default is "%(tag)s %(name)-35.35s %(VR)s: %(repval)s"
indent_format (str or None) – Placeholder for future functionality.

Yields:

str – A string representation of an element.

Return a Dataset from a DICOM JSON Model object.

See the DICOM Standard, Part 18, Annex F.

Parameters:

json_dataset (dict, str, bytes or bytearray) – dict, str, bytes or bytearray representing a DICOM Data Set formatted based on the DICOM JSON Model.
bulk_data_uri_handler (callable, optional) – Callable function that accepts either the tag, vr and “BulkDataURI” value or just the “BulkDataURI” value of the JSON representation of a data element and returns the actual value of that data element (retrieved via DICOMweb WADO-RS). If no bulk_data_uri_handler is specified (default) then the corresponding element will have an “empty” value such as "", b"" or None depending on the vr (i.e. the Value Multiplicity will be 0).

Return type:

Dataset

get(key: str, default: Any | None = None) → Any[source]¶

get(key: int | tuple[int, int] | BaseTag, default: Any | None = None) → DataElement

Simulate dict.get() to handle element tags and keywords.

Parameters:

key (str or int or Tuple[int, int] or BaseTag) – The element keyword or tag or the class attribute name to get.
default (obj or None, optional) – If the element or class attribute is not present, return default (default None).

Returns:

value – If key is the keyword for an element in the Dataset then return the element’s value.
dataelem.DataElement – If key is a tag for a element in the Dataset then return the DataElement instance.
value – If key is a class attribute then return its value.

get_item(key: slice, *, keep_deferred: bool = False) → Dataset[source]¶

get_item(key: int | str | tuple[int, int] | BaseTag, *, keep_deferred: bool = False) → DataElement

Return the raw data element if possible.

It will be raw if the user has never accessed the value, or set their own value. Note if the data element is a deferred-read element, then it is read and converted before being returned.

Parameters:

key – The DICOM (group, element) tag in any form accepted by Tag() such as [0x0010, 0x0010], (0x10, 0x10), 0x00100010, etc. May also be a slice made up of DICOM tags.
keep_deferred (bool, optional) – If True then when returning RawDataElement do not perform the deferred read of the element’s value (accessing the value will return None instead). Default False.

Returns:

The corresponding element.

Return type:

dataelem.DataElement | dataelem.RawDataElement

get_private_item(group: int, element_offset: int, private_creator: str) → DataElement[source]¶

Return the data element for the given private tag group.

This is analogous to Dataset.__getitem__(), but only for private tags. This allows to find the private tag for the correct private creator without the need to add the tag to the private dictionary first.

Parameters:

group (int) – The private tag group where the item is located as a 32-bit int.
element_offset (int) – The lower 16 bits (e.g. 2 hex numbers) of the element tag.
private_creator (str) – The private creator for the tag. Must match the private creator for the tag to be returned.

Returns:

The corresponding element.

Return type:

dataelem.DataElement

Raises:

ValueError – If group is not part of a private tag or private_creator is empty.
KeyError – If the private creator tag is not found in the given group. If the private tag is not found.

group_dataset(group: int) → Dataset[source]¶

Return a Dataset containing only elements of a certain group.

Parameters:: group (int) – The group part of a DICOM (group, element) tag.
Returns:: A Dataset containing elements of the group specified.
Return type:: Dataset

property is_decompressed: bool¶: Return True if the dataset uses an uncompressed (native) transfer syntax, False otherwise.

Added in version 3.1.

property is_implicit_VR: bool | None¶

Get/set the VR method used when encoding the dataset.

Deprecated since version 3.0: is_implicit_VR will be removed in v4.0, set the Transfer Syntax UID or use the implicit_vr argument with save_as() or dcmwrite() instead.

Returns:: If the dataset has been created from scratch then returns None, otherwise returns the VR encoding method used by the decoded dataset.
Return type:: bool | None

property is_little_endian: bool | None¶

Get/set the endianness used when encoding the dataset.

Deprecated since version 3.0: is_little_endian will be removed in v4.0, set the Transfer Syntax UID or use the little_endian argument with save_as() or dcmwrite() instead.

Returns:: If the dataset has been created from scratch then returns None, otherwise returns the endianness of the encoding used by the decoded dataset.
Return type:: bool | None

property is_original_encoding: bool¶

Return True if the encoding to be used for writing is set and is the same as that used to originally encode the Dataset.

This includes properties related to endianness, VR handling and the (0008,0005) Specific Character Set.

items() → Set[tuple[BaseTag, DataElement | RawDataElement]][source]¶

Return the Dataset items to simulate dict.items().

Returns:: The top-level (BaseTag, DataElement) items for the Dataset.
Return type:: dict_items

iterall() → Iterator[DataElement][source]¶

Iterate through the Dataset, yielding all the elements.

Unlike iter(Dataset), this does recurse into sequences, and so yields all elements as if dataset were “flattened”.

Yields:: dataelem.DataElement

keys() → Set[BaseTag][source]¶

Return the Dataset keys to simulate dict.keys().

Returns:: The BaseTag of all the elements in the Dataset.
Return type:: dict_keys

property original_character_set: str | MutableSequence[str]¶

Return the original character set encoding for a dataset decoded from a file or buffer.

Returns:: The original character set encoding of the dataset as given by the (0008,0005) Specific Character Set, or iso8859 if the dataset has been created from scratch.
Return type:: str | MutableSequence[str]

property original_encoding: tuple[bool, bool] | tuple[None, None]¶

Return the original encoding used for a dataset decoded from a file or buffer.

Returns:: For a dataset decoded from a file or buffer this is whether the encoding used implicit/explicit VR and little/big endian as (encoded as implicit VR, encoded as little endian). Returns (None, None) for a dataset created from scratch.
Return type:: tuple[bool, bool] | tuple[None, None]

overlay_array(group: int) → ndarray[source]¶

Return the Overlay Data in group as a numpy.ndarray.

Parameters:: group (int) – The group number of the overlay data.
Returns:: The (group,3000) Overlay Data converted to a numpy.ndarray.
Return type:: numpy.ndarray

property pixel_array: ndarray¶

Return the pixel data as a numpy.ndarray.

Warning

This property requires NumPy and may require the installation of additional packages to perform the actual pixel data decoding. See the pixel data decompression documentation for more information.

Changed in version 3.0: The backend used for pixel data decoding has changed from the pixel_data_handlers module to the pixels module. The behavior of the new backend is not backwards compatible with the old one, in particular the default color space should now be RGB when previously YCbCr data was returned.

To revert to the deprecated pixel_data_handlers backend pass use_v2_backend=True to the pixel_array_options() method:

>>> from pydicom import examples
>>> ds = examples.ct
>>> ds.pixel_array_options(use_v2_backend=True)
>>> arr = ds.pixel_array

The pixel_data_handlers module and the use_v2_backend keyword argument will be removed in v4.0.

Returns:

The contents of the (7FE0,0008) Float Pixel Data, (7FE0,0009) Double Float Pixel Data or (7FE0,0010) Pixel Data elements converted to a numpy.ndarray. The array will be shaped as:

(rows, columns) for single frame, single sample data
(rows, columns, samples) for single frame, multi-sample data
(frames, rows, columns) for multi-frame, single sample data
(frames, rows, columns, samples) for multi-frame, multi-sample data

When using the pydicom.pixels backend the decoding options used with the returned array can be customized via the pixel_array_options() method.

Return type:

numpy.ndarray

See also

pydicom.filewriter.dcmwrite: Encode a Dataset and write it to a file or buffer.

set_original_encoding(is_implicit_vr: bool | None, is_little_endian: bool | None, character_encoding: str | MutableSequence[str] | None = None) → None[source]¶

Set the values for the original dataset encoding.

Can be used for a Dataset with raw data elements to enable optimized writing (e.g. without decoding the data elements).

Changed in version 3.0: character_encoding is now optional

Parameters:

is_implicit_vr (bool | None) – The the original VR encoding of the dataset, True for implicit VR, False for explicit VR or None to reset.
is_little_endian (bool | None) – Set the original endianness of the dataset, True for little endian, False for big or None to reset.
character_encoding (str | MutableSequence[str], optional) – Set the original character set encoding of the dataset. If None then no changes will be made to the original character set encoding.

set_pixel_data(arr: ndarray, photometric_interpretation: str, bits_stored: int, *, generate_instance_uid: bool = True) → None[source]¶

Use an ndarray to set the Pixel Data and related Image Pixel module elements.

Added in version 3.0.

The following Image Pixel module elements values will be added, updated or removed as necessary:

(0028,0002) Samples per Pixel using a value corresponding to photometric_interpretation.
(0028,0104) Photometric Interpretation from photometric_interpretation.
(0028,0006) Planar Configuration will be added and set to 0 if Samples per Pixel is > 1, otherwise it will be removed.
(0028,0008) Number of Frames from the array shape, however it will be removed if arr only contains a single frame.
(0028,0010) Rows and (0028,0011) Columns from the array shape.
(0028,0100) Bits Allocated from the array dtype.
(0028,0101) Bits Stored and (0028,0102) High Bit from bits_stored.
(0028,0103) Pixel Representation from the array dtype.

In addition:

The Transfer Syntax UID will be set to Explicit VR Little Endian if it doesn’t already exist or uses a compressed (encapsulated) transfer syntax.
If generate_instance_uid is True (default) then the SOP Instance UID will be added or updated.
If a bool array is used then the pixel data will be bit-packed using pack_bits().

Changed in version 3.1: Added the ability to use a bool ndarray for Bits Allocated 1 with bit-packing.

Parameters:

arr (numpy.ndarray) –
An array with dtype bool, uint8, uint16, int8, or int16. The array must be shaped as one of the following:
- (rows, columns) for a single frame of grayscale data.
- (frames, rows, columns) for multi-frame grayscale data.
- (rows, columns, samples) for a single frame of multi-sample data such as RGB.
- (frames, rows, columns, samples) for multi-frame, multi-sample data.
photometric_interpretation (str) – The value to use for (0028,0004) Photometric Interpretation. Valid values are "MONOCHROME1", "MONOCHROME2", "PALETTE COLOR", "RGB", "YBR_FULL", "YBR_FULL_422".
bits_stored (int) – The value to use for (0028,0101) Bits Stored. Must be no greater than the number of bits used by the itemsize of arr, or 1 in the case of an array of dtype bool.
generate_instance_uid (bool, optional) – If True (default) then add or update the (0008,0018) SOP Instance UID element with a value generated using generate_uid().

Raises:

NotImplementedError – If the dataset has a big-endian Transfer Syntax UID.

setdefault(key: int | str | tuple[int, int] | BaseTag, default: Any | None = None) → DataElement[source]¶

Emulate dict.setdefault() with support for tags and keywords.

Examples

>>> ds = Dataset()
>>> elem = ds.setdefault((0x0010, 0x0010), "Test")
>>> elem
(0010,0010) Patient's Name                      PN: 'Test'
>>> elem.value
'Test'
>>> elem = ds.setdefault('PatientSex',
...     DataElement(0x00100040, 'CS', 'F'))
>>> elem.value
'F'

Parameters:

key (int, str or 2-tuple of int) –
- If tuple - the group and element number of the DICOM tag
- If int - the combined group/element number
- If str - the DICOM keyword of the tag
default (pydicom.dataelem.DataElement or object, optional) – The DataElement to use with key, or the value of the DataElement to use with key (default None).

Returns:

The DataElement for key.

Return type:

pydicom.dataelem.DataElement or object

Raises:

ValueError – If key is not convertible to a valid tag or a known element keyword.
KeyError – If reading_validation_mode is RAISE and key is an unknown non-private tag.

to_json(bulk_data_threshold: int = 1024, bulk_data_element_handler: Callable[[DataElement], str] | None = None, dump_handler: Callable[[dict[str, Any]], str] | None = None, suppress_invalid_tags: bool = False) → str[source]¶

Return a JSON representation of the Dataset.

See the DICOM Standard, Part 18, Annex F.

Parameters:

bulk_data_threshold (int, optional) – Threshold for the length of a base64-encoded binary data element above which the element should be considered bulk data and the value provided as a URI rather than included inline (default: 1024). Ignored if no bulk data handler is given.
bulk_data_element_handler (callable, optional) – Callable function that accepts a bulk data element and returns a JSON representation of the data element (dictionary including the “vr” key and either the “InlineBinary” or the “BulkDataURI” key).
dump_handler (callable, optional) –
Callable function that accepts a dict and returns the serialized (dumped) JSON string (by default uses json.dumps()).
suppress_invalid_tags (bool, optional) – Flag to specify if errors while serializing tags should be logged and the tag dropped or if the error should be bubbled up.

Returns:

Dataset serialized into a string based on the DICOM JSON Model.

Return type:

str

Examples

>>> def my_json_dumps(data):
...     return json.dumps(data, indent=4, sort_keys=True)
>>> ds.to_json(dump_handler=my_json_dumps)

to_json_dict(bulk_data_threshold: int = 1024, bulk_data_element_handler: Callable[[DataElement], str] | None = None, suppress_invalid_tags: bool = False) → dict[str, Any][source]¶

Return a dictionary representation of the Dataset conforming to the DICOM JSON Model as described in the DICOM Standard, Part 18, Annex F.

Parameters:

bulk_data_threshold (int, optional) – Threshold for the length of a base64-encoded binary data element above which the element should be considered bulk data and the value provided as a URI rather than included inline (default: 1024). Ignored if no bulk data handler is given.
bulk_data_element_handler (callable, optional) – Callable function that accepts a bulk data element and returns a JSON representation of the data element (dictionary including the “vr” key and either the “InlineBinary” or the “BulkDataURI” key).
suppress_invalid_tags (bool, optional) – Flag to specify if errors while serializing tags should be logged and the tag dropped or if the error should be bubbled up.

Returns:

Dataset representation based on the DICOM JSON Model.

Return type:

dict

top() → str[source]¶: Return a str representation of the top level elements.

trait_names() → list[str][source]¶

Return a list of valid names for auto-completion code.

Used in IPython, so that data element names can be found and offered for autocompletion on the IPython command line.

update(d: Dataset | MutableMapping[BaseTag, DataElement | RawDataElement]) → None[source]¶

Extend dict.update() to handle DICOM tags and keywords.

Parameters:: d (dict or Dataset) – The dict or Dataset to use when updating the current object.

Modify the VR or value for the raw element with tag.

When a Dataset is created most of it’s elements are in their RawDataElement form, and only upon trying to access the element is it converted to a DataElement. When this conversion fails due to non-conformance issues, this method can be used to modify the raw element data prior to conversion in order to fix any issues.

Example

Change the VR for the element with tag (0029,1026) before conversion to DataElement.

>>> from pydicom import examples
>>> ds = examples.ct
>>> ds.update_raw_element(0x00291026, vr="US")
>>> elem = ds[0x00291026]  # conversion to DataElement occurs here
>>> type(elem)
<class 'pydicom.dataelem.DataElement'>
>>> elem.VR
"US"

Parameters:

tag (int | str | tuple[int, int] | BaseTag) – The tag for a RawDataElement in the dataset.
vr (str, optional) – Required if value is not used, the value to use for the modified element’s VR, if not used then the existing VR will be kept.
value (bytes, optional) – Required if vr is not used, the value to use for the modified element’s raw encoded value, if not used then the existing value will be kept.

values() → ValuesView[DataElement | RawDataElement][source]¶

Return the Dataset values to simulate dict.values().

Returns:: The DataElements that make up the values of the Dataset.
Return type:: dict_keys

walk(callback: Callable[[Dataset, DataElement], None], recursive: bool = True) → None[source]¶

Iterate through the Dataset's elements and run callback on each.

Visit all elements in the Dataset, possibly recursing into sequences and their items. The callback function is called for each DataElement (including elements with a VR of ‘SQ’). Can be used to perform an operation on certain types of elements.

For example, remove_private_tags() finds all elements with private tags and deletes them.

The elements will be returned in order of increasing tag number within their current Dataset.

Parameters:

callback –
A callable function that takes two arguments:
- a Dataset
- a DataElement belonging to that Dataset
recursive (bool, optional) – Flag to indicate whether to recurse into sequences (default True).

waveform_array(index: int) → ndarray[source]¶

Return an ndarray for the multiplex group at index in the (5400,0100) Waveform Sequence.

Added in version 2.1.

Parameters:: index (int) – The index of the multiplex group to return the array for.
Returns:: The Waveform Data for the multiplex group as an ndarray with shape (samples, channels). If (003A,0210) Channel Sensitivity is present then the values will be in the units specified by the (003A,0211) Channel Sensitivity Units Sequence.
Return type:: numpy.ndarray