mrinufft.extras

Extra utilities for non Cartesian MRI.

This modules notably provides way to estimate sensitivity maps and off-resonance correction.

Most of this module is accessible through the following registry :

• coil sensitivity maps: get_smaps():

  Key	Function
  "low_frequency"	low_frequency(traj, shape, kspace_data, backend, threshold=0.1, density=None, max_iter=10, window_fun='ellipse', blurr_factor=0.0, mask=False)
  "espirit"	espirit(traj, shape, kspace_data, backend, density=None, max_iter=10, calib_width=24, kernel_width=6, thresh=0.02, crop=0.95, decim=1)

• off-resonance correction: get_orc_factorization():

  Key	Function
  "svd"	compute_svd_coefficients(field_map, readout_time, mask, L=-1, n_bins=1024, lazy=False, partial_svd=True)
  "mti"	compute_mti_coefficients(field_map, readout_time, mask, L=-1, n_bins=1024, lazy=False)
  "mfi"	compute_mfi_coefficients(field_map, readout_time, mask, L=9, n_bins=1024, lazy=True)

• least-square optimization methods: get_optimizer():

  Key	Function
  "lsqr"	lsqr(operator, kspace_data, damp=0.0, atol=1e-06, btol=1e-06, conlim=100000000.0, max_iter=100, x0=None, x_init=None, callback=None, progressbar=True)
  "lsmr"	lsmr(operator, kspace_data, damp=0.0, atol=1e-06, btol=1e-06, conlim=100000000.0, max_iter=100, x0=None, x_init=None, callback=None, progressbar=True)
  "cg"	cg(operator, kspace_data, damp=0.0, x0=None, x_init=None, max_iter=10, tol=0.0001, progressbar=True, callback=None)

Tip

You can register your own methods to the registries using the following decorators:

• register_smaps(),

• register_orc(),

• register_optim().

This registry system is also available when using a non-Cartesian Fourier operator

For example:

from mrinufft import get_operator
from mrinufft.extras import register_smaps

# get a non-Cartesian Fourier operator
# with espirit sensitivity maps and off-resonance correction
fourier_op = get_operator("nufft", trajectory=trajectory, smaps="espirit")
fourier_op_orc = fourier_op.with_off_resonance_correction(
    interpolators={"name":"svd", "L":10}
)
# select the least-square solver to use for pseudo-inverse computation.
img = fourier_op_orc.pinv_solver(kspace_data, solver="lsqr")

Custom registered functions can be used as well:

@register_smaps("awesome")
def awesome_smaps(kspace_data, trajectory, **kwargs):
    ...

fourier_op = get_operator("nufft", trajectory=trajectory, smaps="awesome")
# using the function is equivalent:
fourier_op = get_operator("nufft", trajectory=trajectory, smaps=awesome_smaps)

Functions

cartesian_espirit	ESPIRIT algorithm on Cartesian data.
cg	Perform conjugate gradient (CG) optimization for image reconstruction.
coil_compression	Coil compression using principal component analysis on k-space data.
compute_mfi_coefficients	Compute off-resonance correction coefficients using Mixed-Frequency-Interpolator.
compute_mti_coefficients	Compute off-resonance correction coefficients using Mixed Time interpolator (MTI).
compute_svd_coefficients	Compute off-resonance correction coefficients using an SVD.
espirit	ESPIRIT algorithm on non-Cartesian data.
fft	Compute n-dimensional FFT along the last dims axes.
fse_simulation	Perform simple analytical Fast Spin Echo simulation.
get_brainweb_map	Get M0, T1 and T2 parametric maps from a brainweb crisp segmentation.
get_complex_fieldmap_rad	Create a complex-valued field-map in rad/s.
get_optimizer	Get the optimizer function from its name.
get_orc_factorization	Get the orc_factorization function from its name.
get_smaps	Get the smaps function from its name.
ifft	Compute n-dimensional IFFT along the last dims axes.
loss_l2_AHreg	Compute the norm of the residual in the image domain.
loss_l2_reg	Compute the regularized least squares loss for MRI reconstruction.
low_frequency	Calculate low-frequency sensitivity maps.
lsmr	Solve a regularized linear least-squares problem with the LSMR algorithm [1]_.
lsqr	Solve a regularized linear least-squares problem with the LSQR algorithm [1]_.
make_b0map	Make radial B0 map.
make_t2smap	Make homogeneous T2* map.