mrpro.phantoms.brainweb.BrainwebSlices

class mrpro.phantoms.brainweb.BrainwebSlices[source]

Bases: Dataset

Dataset of 2D qMRI parameter slices based on Brainweb dataset.

Dataset of agmented 2D qMRI parameter slices based on the segmentations of the Brainweb dataset [AubertBroche2006].

References

[AubertBroche2006]

Aubert-Broche, B., Griffin, M., Pike, G.B., Evans, A.C., & Collins, D.L. (2006). Twenty New Digital Brain Phantoms for Creation of Validation Image Data Bases. IEEE Transactions on Medical Imaging, 25 (11), 1410-1416. https://doi.org/10.1109/TMI.2006.883453

__init__(folder: str | Path = CACHE_DIR_BRAINWEB, what: Sequence[Literal['r1', 'r2', 'm0', 't1', 't2', 'mask', 'tissueclass', 'skl', 'gry', 'wht', 'csf', 'mrw', 'dura', 'fat', 'fat2', 'mus', 'm-s', 'ves']] = ('m0', 'r1', 'r2'), parameters: Mapping[Literal['skl', 'gry', 'wht', 'csf', 'mrw', 'dura', 'fat', 'fat2', 'mus', 'm-s', 'ves'], BrainwebTissue] = VALUES_3T_RANDOMIZED, orientation: Literal['axial', 'coronal', 'sagittal'] | Sequence[Literal['axial', 'coronal', 'sagittal']] = ('axial', 'coronal', 'sagittal'), skip_slices: tuple[tuple[int, int], tuple[int, int], tuple[int, int]] = ((80, 80), (100, 100), (100, 100)), step: int = 1, slice_preparation: Callable[[Tensor, Generator | None], Tensor] = DEFAULT_AUGMENT_256, mask_values: Mapping[str, float | None] = DEFAULT_VALUES, seed: int | Literal['index', 'random'] = 'random') → None[source]

Initialize Brainweb qMRI slice phantom.

Parameters:

• folder (str | Path, default: CACHE_DIR_BRAINWEB) – Folder with Brainweb data as HDF5 files.

• what (Sequence[Union[Literal['r1', 'r2', 'm0', 't1', 't2', 'mask', 'tissueclass'], Literal['skl', 'gry', 'wht', 'csf', 'mrw', 'dura', 'fat', 'fat2', 'mus', 'm-s', 'ves']]], default: ('m0', 'r1', 'r2')) –

  What to return for each slice. Possible values are:

  • r1: R1 relaxation rate.

  • r2: R2 relaxation rate.

  • m0: M0 magnetization.

  • t1: T1 relaxation time.

  • t2: T2 relaxation time.

  • mask: mask indicating valid data.

  • tissueclass: class index.

• parameters (Mapping[Literal['skl', 'gry', 'wht', 'csf', 'mrw', 'dura', 'fat', 'fat2', 'mus', 'm-s', 'ves'], BrainwebTissue], default: VALUES_3T_RANDOMIZED) –

  Parameters for each tissue class. The Brainweb tissue classes are:

  • skl: skull

  • gry: gray matter

  • wht: white matter

  • csf: cerebrospinal fluid

  • mrw: bone marrow

  • dura: dura

  • fat: fat

  • fat2: fat and Tissue

  • mus: muscle

  • m-s: skin

  • ves: vessels

• orientation (Union[Literal['axial', 'coronal', 'sagittal'], Sequence[Literal['axial', 'coronal', 'sagittal']]], default: ('axial', 'coronal', 'sagittal')) – Orientation of slices (axial, coronal, or sagittal).

• skip_slices (tuple[tuple[int, int], tuple[int, int], tuple[int, int]], default: ((80, 80), (100, 100), (100, 100))) – Specifies how much many slices to skip from the beginning and end for each of axial, coronal, or sagittal orientation.

• step (int, default: 1) – Step size between slices, in voxel units (1 mm).

• slice_preparation (Callable[[Tensor, Generator | None], Tensor], default: DEFAULT_AUGMENT_256) – Callable that performs slice augmentation and resizing, see resize or augment for examples. The default applies slight random rotation, shear, scaling, and flips, and scales to 256x256 images.

• mask_values (Mapping[str, float | None], default: DEFAULT_VALUES) – Default values to use for masked out regions.

• seed (Union[int, Literal['index', 'random']], default: 'random') – Random seed. Can be an int, the strings index to use slice index as seed, or random for random seed.

static download(output_directory: str | PathLike = CACHE_DIR_BRAINWEB, workers: int = 4, progress: bool = False, compress: bool = False) → None[source]

Download Brainweb data.

Parameters:

• output_directory (str | PathLike, default: CACHE_DIR_BRAINWEB) – Directory to save the data.

• workers (int, default: 4) – Number of parallel downloads.

• progress (bool, default: False) – Show progress bar.

• compress (bool, default: False) – Use compression for HDF5 files. Saves disk space but might slow down or speed up access, depending on the file system and access pattern.

__eq__(value, /)

Return self==value.

__new__(**kwargs)