mridataPy

mridataPy is

• a lightweight toolbox for downloading and processing mridata from mridata.org

mridataPy supports to

• download dataset from either mridata.org or old.mridata.org
• load mridata to NumPy arrays, which can be stored to .npy files
• generate sampling masks that can densely sample the center region in k-space while subsample the outer region based on acceleration factor
• provide the ground truth reconstructed by applying RSS coil-combination to fullysampled data
• evaluate reconstructions with MSE, NMSE, PSNR, SSIM metrics

Quickstart

Download one case of Stanford Fullysampled 3D FSE Knees (totally, 20 cases) from mridata.org:

import mridatapy

mridata = mridatapy.data.MRIData()
mridata.download(num=1)

Dependencies and Installation

Package Dependencies

pip will handle all package dependencies.

Install mridataPy

$ pip install mridatapy

Documentation

Module data

MRIData

class mridatapy.data.MRIData(data_type=None, path=None)

• urls

  Attribute: Whole lists of download URLs corresponding to mridata of the given data type.

• filenames

  Attribute: Whole lists of download filenames corresponding to mridata of the given data type.

• type

  Attribute: Data type of mridata.

• dir

  Attribute: Directory to the folder "mridata/" as the default path for mridata.

• download(num=None)

  Instance method: Downloads mridata of the given data type.

• to_np(num=None, stack=None)

  Instance method: Loads mridata to complex-valued k-space NumPy arrays. If not exist, download first.

• to_npy(path=None, num=None, stack=None)

  Instance method: Converts mridata to .npy files. If not exist, download first.

• get(data_type)

  Static method: Gets whole lists of download URLs and filenames corresponding to mridata of the given data type to be downloaded.

• fetch(url, filename, path)

  Static method: Fetches mridata given the specific pair of download URL and filename.

• ismrmrd_to_np(file, filter=None, prewhiten=None, first_slice=None)

  Static method: Loads .h5 ISMRMRD file to complex-valued k-space NumPy array.

• ismrmrd_to_npy(file, path=None, filter=None, prewhiten=None, first_slice=None)

  Static method: Converts .h5 ISMRMRD file to .npy file.

• cfl_to_np(file)

  Static method: Loads .cfl file to complex-valued k-space NumPy array.

• cfl_to_npy(file, path=None)

  Static method: Converts .cfl file to .npy file.

• unzip(file, path=None, remove=None)

  Static method: Unzips .zip file.

• load_npy(file)

  Static method: Loads .npy file.

RandomLine

class mridatapy.data.RandomLine(acceleration_factor, center_fraction)

Generates a sampling mask of the given shape that can densely sample the center region in k-space while subsample the outer region based on acceleration factor. The mask randomly selects a subset of columns from input k-space data.

• __call__(shape, dtype=numpy.complex64, max_attempts=30, tolerance=0.1, seed=None)

  Magic method enables instances to behave like functions.

EquispacedLine

class mridatapy.data.EquispacedLine(acceleration_factor, center_fraction)

Generates a sampling mask of the given shape that can densely sample the center region in k-space while subsample the outer region based on acceleration factor. The mask selects a roughly equispaced subset of columns from input k-space data.

• __call__(shape, dtype=numpy.complex64, max_attempts=30, tolerance=0.1, seed=None)

  Magic method enables instances to behave like functions.

PoissonDisk

class mridatapy.data.PoissonDisk(acceleration_factor, center_fraction)

Generates a sampling mask of the given shape that can densely sample the center region in k-space while subsample the outer region based on acceleration factor. The mask selects a subset of points from input k-space data, characterized by the Poisson disk sampling pattern.

• __call__(shape, dtype=numpy.complex64, max_attempts=30, tolerance=0.1, seed=None)

  Magic method enables instances to behave like functions.

Module utils

fft_centered

function mridatapy.utils.fft_centered(input, shape=None, dim=None, norm=None)

Computes the centered N dimensional discrete Fourier transform (FFT) of input.

ifft_centered

function mridatapy.utils.ifft_centered(input, shape=None, dim=None, norm=None)

Computes the centered N dimensional inverse discrete Fourier transform (IFFT) of input.

root_sum_squares

function mridatapy.utils.root_sum_squares(input, dim, complex=False)

Computes the Root Sum of Squares (RSS) of input along the a given dimension (coil dimension).

Module metrics

mean_squared_error

function mridatapy.metrics.mean_squared_error(gt, pred)

Computes the Mean Squared Error (MSE) between two images.

normalized_mse

function mridatapy.metrics.normalized_mse(gt, pred)

Computes the Normalized Mean Squared Error (NMSE) between two images.

peak_signal_noise_ratio

function mridatapy.metrics.peak_signal_noise_ratio(gt, pred, data_range=None)

Computes the Peak Signal to Noise Ratio (PSNR) between two images.

structural_similarity

function mridatapy.metrics.structural_similarity(gt, pred, data_range=None)

Computes the Structural Similarity Index (SSIM) between two images.

Related Projects

• mridata.org
• mridata-python (pypi.org/project/mridata/)
• SigPy