From BART and sigpy

import bartorch.tools as bt
from bartorch import apps, linop, optim, priors

kspace = bt.phantom(128, coils=8, kspace=True)  # (coils, z, y, x)
maps = bt.ecalib(kspace, maps=1)                # bart ecalib -m1
image = apps.pics(kspace, maps,                 # bart pics -R W:3:0:0.005
                  regularizers=priors.Wavelet((-1, -2), 0.005))

A = linop.CartesianSense(maps, (1, 128, 128))   # sigpy.mri.linop.Sense
x = optim.CG(maxiter=20)(kspace, A)             # sigpy.app.LinearLeastSquares

Arrays are torch tensors in C order, with the coils first: (coils, z, y, x), the reverse of BART’s dimension vector and the layout sigpy uses (Data layout and conventions). Axes are given as indices, never as BART bitmasks.

BART	sigpy	bartorch
bart fft -u -i 7	sp.ifft(x, axes=(-3, -2, -1))	bartorch.ifft(x, (-3, -2, -1), unitary=True)
bart ecalib	mr.app.EspiritCalib	bartorch.tools.ecalib
bart pics -R W:...	mr.app.L1WaveletRecon	apps.pics(..., regularizers=priors.Wavelet(...))
bart pics -R T:...	mr.app.TotalVariationRecon	apps.pics(..., regularizers=priors.TotalVariation(...))
bart pics -t traj	mr.app.SenseRecon(coord=...)	apps.pics(..., traj=traj)
bart nufft	sp.nufft	bartorch.nufft, linop.NUFFT
bart nlinv	—	bartorch.tools.nlinv
bart moba, bart mobafit	—	apps.moba, apps.mobafit
bart traj, bart phantom	mr.radial, sp.shepp_logan	bartorch.tools.traj, bartorch.tools.phantom
—	A.H, A * B, A + B	A.H, A @ B, A + B
—	sp.alg.ConjugateGradient	optim.CG
—	sp.alg.GradientMethod(prox=...)	optim.FISTA, optim.IST
bart <command> ...	—	bartorch <command> ..., the same arguments

BART’s fft is unnormalised unless -u; linop.FFT is unitary. Every other command is a function of bartorch.tools (BART commands as functions).