bartorch.nlop.InversionRecovery

bartorch.nlop.InversionRecovery#

bartorch.nlop.InversionRecovery(TI, shape=(), *, TR=None, bounds=None, unknown=('T1',), amplitude=True, subspace=None, **scale)#

T1 from an inversion recovery: moba -L’s family, on TorchSim.

The longitudinal magnetization read at a series of inversion times, which is what moba’s Look-Locker models fit. moba -L solves for (Mss, M0, R1*) and this solves for T1 and a complex amplitude, so the two agree on the recovery and not on the variables.

Parameters:

  • TI (sequence of float) – Inversion times, in milliseconds.

  • shape (tuple of int) – The voxel shape, C order.

  • TR (float, optional) – Repetition time in milliseconds; without one the magnetization is fully relaxed when each inversion arrives.

  • bounds (dict, optional) – {name: (low, high)}; T1 defaults to (10, 5000) ms, which is what keeps a Gauss-Newton iterate physical.

  • unknown (sequence of str) – What to solve for. inv_efficiency and offset are the other things the model exposes.

  • amplitude (bool) – Carry a complex amplitude multiplying the recovery.

  • subspace (torchsim.Subspace, optional) – Solve in a temporal basis rather than in the contrasts.

  • **scale – The size of a step in a parameter left unbounded.