BART: Features

Figure: Simulated MRI images.

List of Features

• Basic features:
  • runs on Linux and Mac OS X
  • multi-dimensional operations on arrays
  • fast non-uniform Fourier Transform (nuFFT, convolution-based method, GPU gridding)
  • multi-dimensional (divergence-free) wavelet transform
  • parallel computation on multiple cores and with Graphical Processing Units (GPU)
• Iterative methods:
  • Conjugate Gradients (CG)
  • (Fast) Iterative Soft-Thresholding Algorithm (ISTA and FISTA)
  • Normalized iterative hard thresholding (NIHT)
  • Alternating Direction Method of Multipliers (ADMM)
  • Iteratively Regularized Gauss-Newton Method (IRGNM)
  • Primal-dual hybrid gradient algorithm
  • Adam, stochastic gradient descent, adadelta, iPALM
• Calibration methods:
  • direct calibration of coil sensitivities from k-space center
  • Walsh's method for calibration of coil sensitivities
  • ESPIRiT
  • (geometric) channel compression and whitening
  • RING: estimation of gradient delays for radial MRI
• Reconstruction methods for MRI:
  • iterative parallel imaging reconstruction: POCSENSE, SENSE
  • compressed sensing and parallel imaging
  • calibration-less parallel imaging: NLINV and ENLIVE (non-linear optimization) and SAKE (structured low-rank matrix completion)
  • reconstruction with linear subspace constraints
  • non-linear model-based reconstruction for T1, T2, T2*, flow, and water-fat mapping
  • methods based on deep learning: variational networks and MoDL
• Regularization (in arbitrary dimensions):
  • Tikhonov
  • total (generalized) variation
  • l1-wavelet
  • (multi-scale) low-rank
  • Tensorflow-based priors