I/O

input/output management

---

ScalarImage

 ScalarImage (*args, **kwargs)

Image whose pixel values represent scalars.

See :class:~torchio.Image for more information.

Image Writers

---

source

write_image

def write_image(
    data, file_path, dimension_order:str='TCZYX'
):

Writes an image to a file.

:param data: Image data (numpy array, tensor, or AICSImage) :param file_path: Path to save the image :param format: Format to save the image in (default is png)

# Example usage:
numpy_array = np.random.rand(3, 100, 100)
write_image(numpy_array, './data_examples/output_from_numpy.tiff')

tensor = torch.rand(3, 100, 100)
write_image(tensor, './data_examples/output_from_tensor.tiff')

aics_image = AICSImage('./data_examples/example_tiff.tiff')
write_image(aics_image, './data_examples/output_from_tiff.png')

Attempted file (/home/biagio/bioMONAI/nbs/data_examples/example_tiff.tiff) load with reader: <class 'bioio_ome_tiff.reader.Reader'> failed with error: bioio-ome-tiff does not support the image: '/home/biagio/bioMONAI/nbs/data_examples/example_tiff.tiff'. Failed to parse XML for the provided file. Error: no element found: line 1, column 0

Image successfully saved to ./data_examples/output_from_numpy.tiff
Image successfully saved to ./data_examples/output_from_tensor.tiff
Image successfully saved to ./data_examples/output_from_tiff.png

Image Readers

It uses helper functions such as _load_and_preprocess and _multi_sequence to load and preprocess biological images.

---

source

tiff2torch

def tiff2torch(
    file_path:str
):

Load tiff into pytorch tensor

---

source

string2dict

def string2dict(
    input_string:str
):

---

source

split_path

def split_path(
    file_path, # The path to the file to split
    exts:(<class 'fastcore.foundation.L'>, <class 'list'>)=['.ome.tiff', '.tiff', '.tif', '.png'], # List of filename extensions
):

---

source

aics_image_reader

def aics_image_reader(
    ind_dict:NoneType=None, # Dictionary indicating the channels to load
):

Initialize self. See help(type(self)) for accurate signature.

file_path = 'data_examples/example_tiff.tiff'
test_img, _ = aics_image_reader({'Z': 0})(file_path)
test_eq(test_img.shape, (1, 1, 512, 512))

Hierarchical Data Format

---

source

split_hdf_path

def split_hdf_path(
    file_path, # The path to the HDF5 file to split
    hdf5_exts:(<class 'fastcore.foundation.L'>, <class 'list'>)=['.h5', '.hdf5'], # List of filename extensions
):

---

source

hdf5_reader

def hdf5_reader(
    dataset:NoneType=None, # The dataset to load
    patch:int=0, # The patch to load from the dataset
    hdf5_exts:(<class 'fastcore.foundation.L'>, <class 'list'>)=['.h5', '.hdf5'], # List of filename extensions
):

Initialize self. See help(type(self)) for accurate signature.

Images can be loaded by explicitly writing dataset name and path number…

from bioMONAI.visualize import plot_image

file_path = './data_examples/0450_1.hdf5'
dataset_name='clean'
patch_num=10

im , _ = hdf5_reader(dataset=dataset_name, patch=patch_num)(file_path)
plot_image(im[0])

… or enconding them in the path, where datasets are subfolders and patches the image files. The latter being compatible with image_reader syntaxis.

f = file_path + '/' + dataset_name + '/' + '%d'%(patch_num)
im , _ = hdf5_reader()(f)
plot_image(im[0])

Preprocessing

Load and preprocess

org_img, _, _ = _load_and_preprocess(f)

test_eq(org_img.data[0].shape, im.shape)

Read multichannel data

t = _multi_sequence([f], only_tensor=True);
test_eq(t[0].shape, im.shape)

t.shape

torch.Size([1, 1, 96, 96])

Image reader

---

source

image_reader

def image_reader(
    file_path:(<class 'str'>, <class 'pathlib.Path'>, <class 'fastcore.foundation.L'>, <class 'list'>), # Path to the image
    dtype:_TensorMeta=Tensor, # Datatype for the return value. Defaults to torchTensor
    only_tensor:bool=True, # To return only an image tensor
    kwargs:VAR_KEYWORD
):

Loads and preprocesses a medical image.

Args: file_path: Path to the image. Can be a string, Path object or a list. dtype: Datatype for the return value. Defaults to torchTensor. reorder: Whether to reorder the data to be closest to canonical (RAS+) orientation. Defaults to False. resample: Whether to resample image to different voxel sizes and image dimensions. Defaults to None. only_tensor: To return only an image tensor. Defaults to True.

Returns: The preprocessed image. Returns only the image tensor if only_tensor is True, otherwise returns original image, preprocessed image, and original size.

test_eq(image_reader(f)[0].shape, im.shape)