bartorch.tools.sqpics

Contents

bartorch.tools.sqpics#

bartorch.tools.sqpics(kspace, sensitivities, *, l=None, r=None, R=None, s=None, i=None, t=None, n=False, g=False, p=None, I=False, b=None, e=False, H=False, F=False, T=None, W=None, d=None, u=None, C=None, f=None, m=False, w=None, S=False, **extra)#

Parallel-imaging compressed-sensing reconstruction.

Runs bart sqpics.

Parameters:

  • kspace (torch.Tensor) – Input array.

  • sensitivities (torch.Tensor) – Input array.

  • l (str | None) – toggle l1-wavelet or l2 regularization. (-l)

  • r (float | None) – regularization parameter (-r)

  • R (Regularizer | list[Regularizer] | None) – Regularization terms. (-R)

  • s (float | None) – iteration stepsize (-s)

  • i (int | None) – max. number of iterations (-i)

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

  • n (bool) – disable random wavelet cycle spinning (-n)

  • g (bool) – use GPU (-g)

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

  • I (bool) – (select IST) (-I)

  • b (int | None) – Lowrank block size (-b)

  • e (bool) – Scale stepsize based on max. eigenvalue (-e)

  • H (bool) – (hogwild) (-H)

  • F (bool) – (fast) (-F)

  • T (torch.Tensor | None) – (truth file) (-T)

  • W (torch.Tensor | None) – Warm start with <img> (-W)

  • d (int | None) – Debug level (-d)

  • u (float | None) – ADMM rho (-u)

  • C (int | None) – ADMM max. CG iterations (-C)

  • f (float | None) – restrict FOV (-f)

  • m (bool) – Select ADMM (-m)

  • w (float | None) – scaling (-w)

  • S (bool) – Re-scale the image after reconstruction (-S)

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

Returns:

output

Return type:

torch.Tensor

Contents