Multitask MRI Reconstruction & Segmentation Data Classes

class atommic.collections.multitask.rs.data.mrirs_loader.RSMRIDataset(*args: Any, **kwargs: Any)

Bases: MRIDataset

A dataset class for accelerated-MRI reconstruction and MRI segmentation.

Examples

>>> from atommic.collections.multitask.rs.data.mrirs_loader import RSMRIDataset
>>> dataset = RSMRIDataset(root='data/train', sample_rate=0.1)
>>> print(len(dataset))
100
>>> kspace, imspace, coil_sensitivities, mask, initial_prediction, segmentation_labels, attrs, filename,     slice_num = dataset[0]
>>> print(kspace.shape)
np.array([30, 640, 368])

Note

Extends atommic.collections.common.data.MRIDataset.

process_segmentation_labels(segmentation_labels: numpy.ndarray) → numpy.ndarray

Processes segmentation labels to remove, combine, and separate classes.

Parameters

segmentation_labels (np.ndarray) – The segmentation labels. The shape should be (num_slices, height, width) or (height, width).

Returns

The processed segmentation labels.

Return type

np.ndarray

class atommic.collections.multitask.rs.data.mrirs_loader.SKMTEARSMRIDataset(*args: Any, **kwargs: Any)

Bases: RSMRIDataset

Supports the SKM-TEA dataset for multitask accelerated MRI reconstruction and MRI segmentation.

Note

Extends atommic.collections.multitask.rs.data.mrirs_loader.RSMRIDataset.