download_whole_body_ct_female

download_whole_body_ct_female#

download_whole_body_ct_female( load: bool = True, *, high_resolution: bool = False, ) → MultiBlock | str[source]#

Download a CT image of a female subject with 117 segmented anatomic structures.

This dataset is subject 's1380' from the TotalSegmentator dataset, version 2.0.1, available from Zenodo. See the original paper for details:

Jakob Wasserthal et al., TotalSegmentator: Robust Segmentation of 104 Anatomic Structures in CT Images, Radiology, Jul. 2023, doi: https://doi.org/10.1148/ryai.230024.

The dataset is loaded as a MultiBlock with three blocks:

'ct': ImageData with CT data.
'segmentations': MultiBlock with 117 ImageData blocks, each containing a binary segmentation label. The blocks are named by their anatomic structure (for example, 'heart') and are sorted alphabetically. See the examples below for a complete list label names.
'label_map': ImageData with a label map array. The label map is an alternative representation of the segmentation where the masks are combined into a single scalar array.

Note

The label map is not part of the original data source.

Licensed under Creative Commons Attribution 4.0 International.

Added in version 0.45: Three dictionaries are now included with the dataset’s user_dict to map label names to ids and colors:

'names_to_colors' : maps segment names to 8-bit RGB colors.
'names_to_ids' : maps segment names to integer ids used by the label map.
'ids_to_colors' : maps label ids to colors.

The label ids are the ids used by the included label map.

Changed in version 0.45: A down-sampled version of this dataset with dimensions (160, 160, 273) is now returned. Previously, a high-resolution version with dimensions (320, 320, 547) was returned. Use high_resolution=True for the high-resolution version.

Parameters:

loadbool, default: True: Load the dataset after downloading it when True. Set this to False and only the filename will be returned.
high_resolutionbool, default: False: Set this to True to return a high-resolution version of this dataset. By default, a resampled version with a 0.5 sampling rate is returned.

Added in version 0.45.

Returns:

outputpyvista.MultiBlock or str: DataSet or filename depending on load.

Examples#

Download Python source code | Download Jupyter notebook

Load the dataset.

>>> from pyvista import examples
>>> import pyvista as pv
>>> dataset = examples.download_whole_body_ct_female()

Get the names of the dataset’s blocks.

>>> dataset.keys()
['ct', 'segmentations', 'label_map']

Get the CT image.

>>> ct_image = dataset['ct']
>>> ct_image
ImageData (...)
  N Cells:      6825870
  N Points:     6937600
  X Bounds:     7.500e-01, 4.778e+02
  Y Bounds:     7.500e-01, 4.778e+02
  Z Bounds:     7.528e-01, 8.122e+02
  Dimensions:   160, 160, 271
  Spacing:      3.000e+00, 3.000e+00, 3.006e+00
  N Arrays:     1

Get the segmentation label names and show the first three.

>>> segmentations = dataset['segmentations']
>>> label_names = segmentations.keys()
>>> label_names[:3]
['adrenal_gland_left', 'adrenal_gland_right', 'aorta']

Get the label map and show its data range.

>>> label_map = dataset['label_map']
>>> label_map.get_data_range()
(np.uint8(0), np.uint8(117))

Show the 'names_to_colors' dictionary with RGB colors for each segment.

>>> dataset.user_dict['names_to_colors']

Show the 'names_to_ids' dictionary with a mapping from segment names to segment ids.

>>> dataset.user_dict['names_to_ids']

Create a surface mesh of the segmentation labels.

>>> labels_mesh = label_map.contour_labels()

Color the surface using color_labels(). Use the 'ids_to_colors' dictionary included with the dataset to map the colors.

>>> colored_mesh = labels_mesh.color_labels(
...     colors=dataset.user_dict['ids_to_colors']
... )

Plot the CT image and segmentation labels together.

>>> pl = pv.Plotter()
>>> _ = pl.add_volume(
...     ct_image,
...     cmap='bone',
...     opacity='sigmoid_7',
...     show_scalar_bar=False,
... )
>>> _ = pl.add_mesh(colored_mesh)
>>> pl.view_zx()
>>> pl.camera.up = (0, 0, 1)
>>> pl.camera.zoom(1.3)
>>> pl.show()