bartorch.tools.ncalib

Contents

bartorch.tools.ncalib#

bartorch.tools.ncalib(kspace, *, g=False, t=None, p=None, B=None, r=None, i=None, cgiter=None, cgtol=None, alpha=None, M=None, a=None, b=None, c=None, w=None, o=False, N=False, m=None, dims=None, sens_os=None, shared_img_dims=None, shared_col_dims=None, scale_loop_dims=None, phase_pole=None, **extra)#

Estimate coil sensitivities using ENLIVE calibration.

Runs bart ncalib.

Parameters:

  • kspace (torch.Tensor) – Input array.

  • g (bool) – use gpu (-g)

  • t (torch.Tensor | None) – kspace trajectory (-t)

  • p (torch.Tensor | None) – kspace pattern (-p)

  • B (torch.Tensor | None) – subspace basis (-B)

  • r (tuple[int, int, int] | None) – Limits the size of the calibration region. (-r)

  • i (int | None) – Number of Newton steps (-i)

  • cgiter (int | None) – (iterations for linearized problem) (--cgiter)

  • cgtol (float | None) – (tolerance for linearized problem) (--cgtol)

  • alpha (float | None) – (alpha in first iteration) (--alpha)

  • M (float | None) – (minimum for regularization) (-M)

  • a (float | None) – (a in 1 + a * Laplace^-b/2) (-a)

  • b (float | None) – (b in 1 + a * Laplace^-b/2) (-b)

  • c (float | None) – (c in 1 + a * Laplace^-b/2) (-c)

  • w (float | None) – (inverse scaling of the data) (-w)

  • o (bool) – return oversampled coils (-o)

  • N (bool) – Normalize coil sensitivities (-N)

  • m (int | None) – Number of ENLIVE maps to use in reconstruction (-m)

  • dims (tuple[int, int, int] | None) – Explicitly specify sens dimensions (-x)

  • sens_os (float | None) – (over-sampling factor for sensitivities) (--sens-os)

  • shared_img_dims (int | tuple[int, ...] | None) – Axes the image is shared along. (--shared-img-dims)

  • shared_col_dims (int | tuple[int, ...] | None) – Axes the coil sensitivities are shared along. (--shared-col-dims)

  • scale_loop_dims (int | tuple[int, ...] | None) – Scale the parameters as if ncalib were looped over these axes. (--scale-loop-dims)

  • phase_pole (int | None) – Use phase pole detection after d iterations (0 for every iteration) (--phase-pole)

  • **extra (Any) – Further BART flags, passed through by name.

Returns:

sensitivities, image (roughly scaled to rss of lowres k-space)

Return type:

tuple of torch.Tensor

Contents