bartorch.tools.wshfl

Contents

bartorch.tools.wshfl#

bartorch.tools.wshfl(maps, wave, phi, reorder, table, *, R=None, b=None, i=None, j=None, s=None, e=None, F=None, O=None, t=None, g=False, K=False, H=False, v=False, **extra)#

Perform a wave-shuffling reconstruction. Conventions: * (sx, sy, sz) - Spatial dimensions. * wx - Extended FOV in READ_DIM due to wave’s voxel spreading. * (nc, md) - Number of channels and ESPIRiT’s extended-SENSE model operator dimensions (or # of maps). * (tf, tk) - Turbo-factor and the rank of the temporal basis used in shuffling. * ntr - Number of TRs, or the number of (ky, kz) points acquired of one echo image. * n - Total number of (ky, kz) points acquired. This is equal to the product of ntr and tf. Descriptions: * reorder is an (n by 3) index matrix such that [ky, kz, t] = reorder(i, :) represents the (ky, kz) kspace position of the readout line acquired at echo number (t), and 0 <= ky < sy, 0 <= kz < sz, 0 <= t < tf). * table is a (wx by nc by n) matrix such that table(:, :, k) represents the kth multichannel kspace line. Expected dimensions: * maps - ( sx, sy, sz, nc, md, 1, 1) * wave - ( wx, sy, sz, 1, 1, 1, 1) * phi - ( 1, 1, 1, 1, 1, tf, tk) * output - ( sx, sy, sz, 1, md, 1, tk) * reorder - ( n, 3, 1, 1, 1, 1, 1) * table - ( wx, nc, n, 1, 1, 1, 1)

Runs bart wshfl.

Parameters:

  • maps (torch.Tensor) – Input array.

  • wave (torch.Tensor) – Input array.

  • phi (torch.Tensor) – Input array.

  • reorder (torch.Tensor) – Input array.

  • table (torch.Tensor) – Input array.

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

  • b (int | None) – Block size for locally low rank. (-b)

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

  • j (int | None) – Maximum number of CG iterations in ADMM. (-j)

  • s (float | None) – ADMM Rho value. (-s)

  • e (float | None) – Eigenvalue to scale step size. (Optional.) (-e)

  • F (torch.Tensor | None) – Go from shfl-coeffs to data-table. Pass in coeffs path. (-F)

  • O (torch.Tensor | None) – Initialize reconstruction with guess. (-O)

  • t (float | None) – Tolerance convergence condition for FISTA. (-t)

  • g (bool) – Use GPU. (-g)

  • K (bool) – Go from data-table to shuffling basis k-space. (-K)

  • H (bool) – Use hogwild. (-H)

  • v (bool) – Split coefficients to real and imaginary components. (-v)

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

Returns:

output

Return type:

torch.Tensor

Contents