tools.nlinv#

bartorch.tools.nlinv()#

Nonlinear inversion: the image and the sensitivities together.

Parameters:

kspace (torch.Tensor) – Under-sampled k-space, C order.
maxiter (int, default=None) – Gauss-Newton steps (-i).
maps (int, default=None) – How many sets of sensitivities to estimate (-m).
traj (tensor, default=None) – Non-Cartesian trajectory (-t).
pattern (tensor, default=None) – Sampling pattern (-p).
basis (tensor, default=None) – Subspace basis (-B).
initial (tensor, default=None) – Warm start (-I).
alpha (float, default=None) – The a of the Sobolev coil weighting (1 + a |k|^2)^(-b/2) (-a), not the first step’s regularization weight – that is nlinv --alpha, reachable through **extra.
real (bool, default=False) – Constrain the image to be real (-c).
normalize (bool, default=True) – Divide the image by the root sum of squares of the sensitivities, as nlinv does unless told not to. BART spells this the other way round, as -N for “do not normalize”.
return_sensitivities (bool, default=False) – Also return the sensitivities, which BART writes as a second array.
**extra – Further BART nlinv options, by name. s, the axes the sensitivities are constant along, takes axes of kspace.

Returns:

The image, and the sensitivities when asked for.

Return type:

torch.Tensor or tuple of torch.Tensor

Examples using `nlinv`#