AUTOGENERATED. DO NOT EDIT.


--avg--

Usage: avg [-w] bitmask <input> <output> 

Calculates (weighted) average along dimensions specified by bitmask.

-w    weighted average
-h    help


--bench--

Usage: bench [-T] [-S] [-s d] [<output>] 

Performs a series of micro-benchmarks.

-T          varying number of threads
-S          varying problem size
-s flags    select benchmarks
-h          help


--bin--

Usage: bin [-l d] [-o] [-R d] [-C d] [-a d] [-O f:f] [-M] <label> <src> <dst> 

Binning

-l dim         Bin according to labels: Specify cluster dimension
-o             Reorder according to labels
-R n_resp      Quadrature Binning: Number of respiratory labels
-C n_card      Quadrature Binning: Number of cardiac labels
-a window      Quadrature Binning: Moving average
-O [r:c]deg    Quadrature Binning: Angle offset for resp and card.
-M             Amplitude binning
-h             help


--bitmask--

Usage: bitmask [-b] [dim1 ... dimN ] 

Convert between a bitmask and set of dimensions.

-b    dimensions from bitmask, use with exactly one argument
-h    help


--cabs--

Usage: cabs <input> <output> 

Absolute value of array (|<input>|).

-h  help


--calc--

Usage: calc [-L...] func <input> <output> 

Perform function evaluation on array.

-L...    Print a list of all supported functions
-h       help


--caldir--

Usage: caldir cal_size <input> <output> 

Estimates coil sensitivities from the k-space center using
a direct method (McKenzie et al.). The size of the fully-sampled
calibration region is automatically determined but limited by
{cal_size} (e.g. in the readout direction).

-h  help


--calmat--

Usage: calmat [-k d:d:d] [-r d:d:d] <kspace> <calibration_matrix> 

Compute calibration matrix.

-k ksize       kernel size
-r cal_size    Limits the size of the calibration region.
-h             help


--carg--

Usage: carg <input> <output> 

Argument (phase angle).

-h  help


--casorati--

Usage: casorati dim1 kern1 ... dimN kernN <input> <output> 

Casorati matrix with kernel (kern1, ..., kernN) along dimensions (dim1, ..., dimN).

-h  help


--cc--

Usage: cc [-p d] [-M] [-r d:d:d] [-A] [-S] [-G] [-E] <kspace> <coeff|proj_kspace> 

Performs coil compression.

-p N    perform compression to N virtual channels
-M      output compression matrix
-r S    size of calibration region
-A      use all data to compute coefficients
-S      type: SVD
-G      type: Geometric
-E      type: ESPIRiT
-h      help


--ccapply--

Usage: ccapply [-p d] [-u] [-t] [-S] [-G] [-E] <kspace> <cc_matrix> <proj_kspace> 

Apply coil compression forward/inverse operation.

-p N    perform compression to N virtual channels
-u      apply inverse operation
-t      don't apply FFT in readout
-S      type: SVD
-G      type: Geometric
-E      type: ESPIRiT
-h      help


--cdf97--

Usage: cdf97 [-i] bitmask <input> <output> 

Perform a wavelet (cdf97) transform.

-i    inverse
-h    help


--circshift--

Usage: circshift dim shift <input> <output> 

Perform circular shift along {dim} by {shift} elements.

-h  help


--conj--

Usage: conj <input> <output> 

Compute complex conjugate.

-h  help


--conv--

Usage: conv bitmask <input> <kernel> <output> 

Performs a convolution along selected dimensions.

-h  help


--conway--

Usage: conway [-P] [-n d] <input> <output> 

Conway's game of life.

-P      periodic boundary conditions
-n #    nr. of iterations
-h      help


--copy--

Usage: copy [dim1 pos1 ... dimN posN ] <input> <output> 

Copy an array (to a given position in the output file - which then must exist).

-h  help


--cpyphs--

Usage: cpyphs <input> <output> 

Copy phase from <input> to <output>.

-h  help


--creal--

Usage: creal <input> <output> 

Real value.

-h  help


--crop--

Usage: crop dimension size <input> <output> 

Extracts a sub-array corresponding to the central part of {size} along {dimension}

-h  help


--delta--

Usage: delta dims flags size <out> 

Kronecker delta.

-h  help


--ecalib--

Usage: ecalib [-t f] [-c f] [-k d:d:d] [-r d:d:d] [-m d] [-S] [-W] [-I] [-1] [-P] [-v f] [-a] [-d d] <kspace> <sensitivities> [<ev-maps>] 

Estimate coil sensitivities using ESPIRiT calibration.
Optionally outputs the eigenvalue maps.

-t threshold     This determined the size of the null-space.
-c crop_value    Crop the sensitivities if the eigenvalue is smaller than {crop_value}.
-k ksize         kernel size
-r cal_size      Limits the size of the calibration region.
-m maps          Number of maps to compute.
-S               create maps with smooth transitions (Soft-SENSE).
-W               soft-weighting of the singular vectors.
-I               intensity correction
-1               perform only first part of the calibration
-P               Do not rotate the phase with respect to the first principal component
-v variance      Variance of noise in data.
-a               Automatically pick thresholds.
-d level         Debug level
-h               help


--ecaltwo--

Usage: ecaltwo [-c f] [-m d] [-S] x y z <input> <sensitivities> [<ev-maps>] 

Second part of ESPIRiT calibration.
Optionally outputs the eigenvalue maps.

-c crop_value    Crop the sensitivities if the eigenvalue is smaller than {crop_value}.
-m maps          Number of maps to compute.
-S               Create maps with smooth transitions (Soft-SENSE).
-h               help


--epg--

Usage: epg [-C] [-M] [-H] [-F] [-S] [-B] [-1 f] [-2 f] [-b f] [-o f] [-r f] [-e f] [-f f] [-s d] [-n d] [-u d] [-v d] <signal intensity> [<configuration states>] [<(rel.) signal derivatives>] [<configuration derivatives>] 

Simulate MR pulse sequence based on Extended Phase Graphs (EPG)

-C        CPMG
-M        fmSSFP
-H        Hyperecho
-F        FLASH
-S        Spinecho
-B        bSSFP
-1 T1     T1 [units of time]
-2 T2     T2 [units of time]
-b B1     relative B1 [unitless]
-o OFF    off-resonance [units of inverse time]
-r TR     repetition time [units of time]
-e TE     echo time [units of time]
-f FA     flip angle [degrees]
-s SP     spoiling (0: ideal, 1: conventional RF, 2: random RF)
-n N      number of pulses
-u U      unknowns as bitmask (0: T1, 1: T2, 2: B1, 3: off-res)
-v V      verbosity level
-h        help


--estdelay--

Usage: estdelay [-R] [-p d] [-n d] [-r f] <trajectory> <data> [<qf>] 

Estimate gradient delays from radial data.

-R      RING method
-p p    [RING] Padding
-n n    [RING] Number of intersecting spokes
-r r    [RING] Central region size
-h      help


--estdims--

Usage: estdims <traj> 

Estimate image dimension from non-Cartesian trajectory.
Assume trajectory scaled to -DIM/2 to DIM/2 (ie dk=1/FOV=1)

-h  help


--estshift--

Usage: estshift flags <arg1> <arg2> 

Estimate sub-pixel shift.

-h  help


--estvar--

Usage: estvar [-k d:d:d] [-r d:d:d] <kspace> 

Estimate the noise variance assuming white Gaussian noise.

-k ksize       kernel size
-r cal_size    Limits the size of the calibration region.
-h             help


--extract--

Usage: extract dim1 start1 end1 ... dimN startN endN <input> <output> 

Extracts a sub-array along dims from index start to (not including) end.

-h  help


--fakeksp--

Usage: fakeksp [-r] <image> <kspace> <sens> <output> 

Recreate k-space from image and sensitivities.

-r    replace measured samples with original values
-h    help


--fft--

Usage: fft [-u] [-i] [-n] bitmask <input> <output> 

Performs a fast Fourier transform (FFT) along selected dimensions.

-u    unitary
-i    inverse
-n    un-centered
-h    help


--fftmod--

Usage: fftmod [-i] bitmask <input> <output> 

Apply 1 -1 modulation along dimensions selected by the {bitmask}.

-i    inverse
-h    help


--fftrot--

Usage: fftrot dim1 dim2 theta <input> <output> 

Performs a rotation using Fourier transform (FFT) along selected dimensions.

-h  help


--fftshift--

Usage: fftshift [-b] bitmask <input> <output> 

Apply fftshift along dimensions selected by the {bitmask}.

-b    apply ifftshift
-h    help


--filter--

Usage: filter [-m d] [-l d] [-G] [-a d] <input> <output> 

Apply filter.

-m dim    median filter along dimension dim
-l len    length of filter
-G        geometric median
-a dim    Moving average filter along dimension dim
-h        help


--flatten--

Usage: flatten <input> <output> 

Flatten array to one dimension.

-h  help


--flip--

Usage: flip bitmask <input> <output> 

Flip (reverse) dimensions specified by the {bitmask}.

-h  help


--fmac--

Usage: fmac [-A] [-C] [-s d] <input1> [<input2>] <output> 

Multiply <input1> and <input2> and accumulate in <output>.
If <input2> is not specified, assume all-ones.

-A      add to existing output (instead of overwriting)
-C      conjugate input2
-s b    squash dimensions selected by bitmask b
-h      help


--fovshift--

Usage: fovshift [-t <file>] [-s f:f:f] <input> <output> 

Shifts FOV.

-t file     k-space trajectory
-s X:Y:Z    FOV shift
-h          help


--homodyne--

Usage: homodyne [-r f] [-I] [-C] [-P <file>] [-n] dim fraction <input> <output> 

Perform homodyne reconstruction along dimension dim.

-r alpha         Offset of ramp filter, between 0 and 1. alpha=0 is a full ramp, alpha=1 is a horizontal line
-I               Input is in image domain
-C               Clear unacquired portion of kspace
-P phase_ref>    Use <phase_ref> as phase reference
-n               use uncentered ffts
-h               help


--ictv--

Usage: ictv [-i d] [-u f] lambda flags flags <input> <output> 

Infimal convolution of total variation along dims specified by flags.

-i i      max. iterations
-u rho    rho in ADMM
-h        help


--index--

Usage: index dim size <name> 

Create an array counting from 0 to {size-1} in dimensions {dim}.

-h  help


--invert--

Usage: invert <input> <output> 

Invert array (1 / <input>). The output is set to zero in case of divide by zero.

-h  help


--itsense--

Usage: itsense alpha <sensitivities> <kspace> <pattern> <output> 

A simplified implementation of iterative sense reconstruction
with l2-regularization.

-h  help


--join--

Usage: join [-a] dimension <input>1> ... <input>N> <output> 

Join input files along {dimensions}. All other dimensions must have the same size.
	 Example 1: join 0 slice_001 slice_002 slice_003 full_data
	 Example 2: join 0 `seq -f "slice_%%03g" 0 255` full_data

-a    append - only works for cfl files!
-h    help


--looklocker--

Usage: looklocker [-t f] [-D f] <input> <output> 

Compute T1 map from M_0, M_ss, and R_1*.

-t threshold    Pixels with M0 values smaller than {threshold} are set to zero.
-D delay        Time between the middle of inversion pulse and the first excitation.
-h              help


--lrmatrix--

Usage: lrmatrix [-d] [-i d] [-m d] [-f d] [-j d] [-k d] [-N] [-s] [-l d] [-o <file>] <input> <output> 

Perform (multi-scale) low rank matrix completion

-d          perform decomposition instead, ie fully sampled
-i iter     maximum iterations.
-m flags    which dimensions are reshaped to matrix columns.
-f flags    which dimensions to perform multi-scale partition.
-j scale    block size scaling from one scale to the next one.
-k size     smallest block size
-N          add noise scale to account for Gaussian noise.
-s          perform low rank + sparse matrix completion.
-l size     perform locally low rank soft thresholding with specified block size.
-o out2     summed over all non-noise scales to create a denoised output.
-h          help


--mandelbrot--

Usage: mandelbrot [-s d] [-n d] [-t f] [-z f] [-r f] [-i f] <output> 

Compute mandelbrot set.

-s size    image size
-n #       nr. of iterations
-t t       threshold for divergence
-z z       zoom
-r r       offset real
-i i       offset imag
-h         help


--measure--

Usage: measure [--mse] [--mse-mag] [--ssim] [--psnr] <reference> <input> [<output>] 



--mse        mse
--mse-mag    mse of rss (over coil dim)
--ssim       ssim of rss (over coil dim) and mean over other dims
--psnr       psnr of rss (over coil dim) and mean over other dims
-h           help


--mip--

Usage: mip [-m] [-a] bitmask <input> <output> 

Maximum (minimum) intensity projection (MIP) along dimensions specified by bitmask.

-m    minimum
-a    do absolute value first
-h    help


--mnist--

Usage: mnist [-a,--apply] [-t,--train] [-g,--gpu] <input> <weights> <ref/output> 

Trains or applies a MNIST network.
This network is to demonstrate how a neural network can be implemented in BART.

-a,--apply    apply nnet
-t,--train    trains network
-g,--gpu      run on gpu
-h            help


--moba--

Usage: moba [-r ...] [-L] [-P] [-F] [-G] [--bloch] [-m d] [-l d] [-i d] [-R,--reduction f] [-T f] [-j f] [-u f] [-C d] [-s f] [-B f] [-b f:f] [-d d] [-f f] [-p <file>] [-J] [-M] [-g] [--positive-maps d] [--not-wav-maps d] [--l2-on-parameters d] [--pusteps d] [--ratio f] [--l1val f] [-I <file>] [-t <file>] [-o f] [--img_dims d:d:d] [-k] [--kfilter-1] [--kfilter-2] [-e f] [--fat_spec_0] [--scale_data f] [--seq ...] [--sim ...] [--other ...] <kspace> <TI/TE> <output> [<sensitivities>] 

Model-based nonlinear inverse reconstruction

-r <T>:A:B:C               generalized regularization options (-rh for help)
-L                         T1 mapping using model-based look-locker
-P                         T1 mapping using reparameterized (M0, R1, alpha) model-based look-locker (TR required!)
-F                         T2 mapping using model-based Fast Spin Echo
-G                         T2* mapping using model-based multiple gradient echo
--bloch                    Bloch model-based reconstruction
-m model                   Select the MGRE model from enum { WF = 0, WFR2S, WF2R2S, R2S, PHASEDIFF } [default: WFR2S]
-l 1/-l2                    toggle l1-wavelet or l2 regularization.
-i iter                    Number of Newton steps
-R,--reduction redu        reduction factor
-T damp                    damping on temporal frames
-j minreg                  Minimum regularization parameter
-u rho                     ADMM rho [default: 0.01]
-C iter                    inner iterations
-s step                    step size
-B bound                   lower bound for relaxation
-b SMO:SC                  B0 field: spatial smooth level; scaling [default: 222.; 1.]
-d level                   Debug level
-f FOV                     
-p PSF                     
-J                         Stack frames for joint recon
-M                         Simultaneous Multi-Slice reconstruction
-g                         use gpu
--positive-maps flag       Maps with positivity contraint as FLAG!
--not-wav-maps d           Maps removed from wavelet denoising (counted from back!)
--l2-on-parameters flag    Flag for parameter maps with l2 norm
--pusteps ud               Number of partial update steps for IRGNM
--ratio f:[0;1]            Ratio of partial updates: ratio*<updated-map> + (1-ratio)*<previous-map>
--l1val f                  Regularization scaling of l1 wavelet (default: 1.)
-I init                    File for initialization
-t traj                    K-space trajectory
-o os                      Oversampling factor for gridding [default: 1.]
--img_dims x:y:z           dimensions
-k                         k-space edge filter for non-Cartesian trajectories
--kfilter-1                k-space edge filter 1
--kfilter-2                k-space edge filter 2
-e kfilter_strength        strength for k-space edge filter [default: 2e-3]
--fat_spec_0               select fat spectrum from ISMRM fat-water tool
--scale_data f             scaling factor for data
--seq ...                  configure sequence parameters
--sim ...                  configure simulation parameters
--other ...                configure other parameters
-h                         help


--mobafit--

Usage: mobafit [-T] [-I] [-L] [-G] [-D] [-m d] [-a] [-i d] [-g] [-B <file>] [--init [f:]*f] [--scale [f:]*f] [--min-flag d] [--max-flag d] [--max-mag-flag d] [--min [f:]*f] [--max [f:]*f] <enc> <echo/contrast images> [<coefficients>] 

Pixel-wise fitting of physical signal models.

-T                      TSE
-I                      Inversion Recovery: f(M0, R1, c) =  M0 * (1 - exp(-t * R1 + c))
-L                      Inversion Recovery Look-Locker
-G                      MGRE
-D                      diffusion
-m model                Select the MGRE model from enum { WF = 0, WFR2S, WF2R2S, R2S, PHASEDIFF } [default: WFR2S]
-a                      fit magnitude of signal model to data
-i iter                 Number of IRGNM steps
-g                      use gpu
-B file                 temporal (or other) basis
--init [f:]*f           Initial values of parameters in model-based reconstruction
--scale [f:]*f          Scaling
--min-flag flags        Apply minimum constraint on selected maps
--max-flag flags        Apply maximum constraint on selected maps
--max-mag-flag flags    Apply maximum magnitude constraint on selected maps
--min [f:]*f            Min bound (map must be selected with "min-flag")
--max [f:]*f            Max bound (map must be selected with "max-flag" or "max-mag-flag")
-h                      help


--morphop--

Usage: morphop [-e] [-d] [-o] [-c] mask_size <binary input> [<binary output>] 

Perform morphological operators on binary data with odd mask sizes.

-e    EROSION (default)
-d    DILATION
-o    OPENING
-c    CLOSING
-h    help


--multicfl--

Usage: multicfl [-s] <cfl>1> ... <cfl>N> 

Combine/Split multiple cfl files to one multi-cfl file.
In normal usage, the last argument is the combined multi-cfl,
with '-s', the first argument is the multi-cfl that is split up

-s    separate
-h    help


--nlinv--

Usage: nlinv [-i d] [-d d] [-c] [-N] [-m d] [-U] [-f f] [-p <file>] [-t <file>] [-I <file>] [-g] [-S] [--lowmem] [-x d:d:d] <kspace> <output> [<sensitivities>] 

Jointly estimate image and sensitivities with nonlinear
inversion using {iter} iteration steps. Optionally outputs
the sensitivities.

-i iter     Number of Newton steps
-d level    Debug level
-c          Real-value constraint
-N          Do not normalize image with coil sensitivities
-m nmaps    Number of ENLIVE maps to use in reconstruction
-U          Do not combine ENLIVE maps in output
-f FOV      restrict FOV
-p file     pattern / transfer function
-t file     kspace trajectory
-I file     File for initialization
-g          use gpu
-S          Re-scale image after reconstruction
--lowmem    Use low-mem mode of the nuFFT
-x x:y:z    Explicitly specify image dimensions
-h          help


--nlmeans--

Usage: nlmeans [-p,--patch_length d] [-d,--patch_dist d] [-H f] [-a f] flags <input> <output> 

Non-local means filter

-p,--patch_length int    patch length
-d,--patch_dist int      patch distance
-H h                     NLMeans h
-a a                     NLMeans a (stddev for gaussian euclidean distance)
-h                       help


--nnet--

Usage: nnet [-a,--apply] [-e,--eval] [-t,--train] [-g,--gpu] [-b,--batch-size d] [-l,--load <file>] [-N,--network ...] [-U,--unet-segm ...] [--train-loss ...] [--valid-loss ...] [--valid-data ...] [-T,--train-algo ...] [--adam ...] [--load-memory] [--export-graph <string>] <input> <weights> <ref/output> 

Trains or applies a neural network.

-a,--apply                   apply nnet
-e,--eval                    evaluate nnet
-t,--train                   trains network
-g,--gpu                     run on gpu
-b,--batch-size batchsize    size of mini batches
-l,--load <weights-init>     load weights for continuing training
-N,--network ...             select neural network
-U,--unet-segm ...           configure U-Net for segmentation
--train-loss ...             configure the training loss
--valid-loss ...             configure the validation loss
--valid-data ...             provide validation data
-T,--train-algo ...          configure general training parmeters
--adam ...                   configure Adam
--load-memory                load files into memory
--export-graph <file.dot>    export graph for visualization
-h                           help


--noise--

Usage: noise [-s d] [-r] [-n f] <input> <output> 

Add noise with selected variance to input.

-s d           random seed initialization
-r             real-valued input
-n variance    DEFAULT: 1.0
-h             help


--normalize--

Usage: normalize [-b] flags <input> <output> 

Normalize along selected dimensions.

-b    l1
-h    help


--nrmse--

Usage: nrmse [-t f] [-s] [-S,--scientific] <reference> <input> 

Output normalized root mean square error (NRMSE),
i.e. norm(input - ref) / norm(ref)

-t eps             compare to eps
-s                 automatic (complex) scaling
-S,--scientific    use scientific notation in output
-h                 help


--nufft--

Usage: nufft [-a] [-i] [-x d:d:d] [-t] [-r] [-c] [-l f] [-m d] [-P] [-s] [-g] [-1] [--lowmem] [--zero-mem] [--no-precomp] [-B <file>] [-p <file>] [-o,--oversampling f] [-w,--width f] <traj> <input> <output> 

Perform non-uniform Fast Fourier Transform.

-a                     adjoint
-i                     inverse
-x x:y:z               dimensions
-t                     Toeplitz embedding for inverse NUFFT
-r                     turn-off Toeplitz embedding for inverse NUFFT
-c                     Preconditioning for inverse NUFFT
-l lambda              l2 regularization
-m iter                max. number of iterations (inverse only)
-P                     periodic k-space
-s                     DFT
-g                     GPU
-1                     use/return oversampled grid
--lowmem               Use low-mem mode of the nuFFT
--zero-mem             Use zero-overhead mode of the nuFFT
--no-precomp           Use low-low-mem mode of the nuFFT
-B file                temporal (or other) basis
-p file                weighting of nufft
-o,--oversampling o    oversample grid by factor (default: o=2; required for Toeplitz)
-w,--width w           width of Kaiser-Bessel window (default: w=6)
-h                     help


--nufftbase--

Usage: nufftbase dimensions <trajectory> <output> 

Compute the Fourier transform of a basis function to be used in the nuFFT.

-h  help


--onehotenc--

Usage: onehotenc [-r] [-i d] <input> <output> 

Transforms class labels to one-hot-encoded classes


-r          get class label by maximum entry
-i index    select dimension
-h          help


--ones--

Usage: ones dims dim1 ... dimN <output> 

Create an array filled with ones with {dims} dimensions of size {dim1} to {dimn}.

-h  help


--pattern--

Usage: pattern [-s d] <kspace> <pattern> 

Compute sampling pattern from kspace

-s bitmask    Squash dimensions selected by bitmask
-h            help


--phantom--

Usage: phantom [-s d] [-S d] [-k] [-t <file>] [-G] [-T] [--NIST] [--SONAR] [--BRAIN] [--ELLIPSOID] [--ellipsoid_center d:d:d] [--ellipsoid_axes f:f:f] [-N d] [-B] [--FILE <file>] [-x d] [-g d] [-3] [-b] [-r d] [--rotation-angle f] [--rotation-steps d] [--coil ...] <output> 

Image and k-space domain phantoms.

-s nc                       nc sensitivities
-S nc                       Output nc sensitivities
-k                          k-space
-t file                     trajectory
-G                          geometric object phantom
-T                          tubes phantom
--NIST                      NIST phantom (T2 sphere)
--SONAR                     Diagnostic Sonar phantom
--BRAIN                     BRAIN geometry phantom
--ELLIPSOID                 Ellipsoid.
--ellipsoid_center d:d:d    x,y,z center coordinates of ellipsoid.
--ellipsoid_axes f:f:f      Axes lengths of ellipsoid.
-N num                      Random tubes phantom with num tubes
-B                          BART logo
--FILE name                 Arbitrary geometry based on multicfl file.
-x n                        dimensions in y and z
-g n=1,2,3                  select geometry for object phantom
-3                          3D
-b                          basis functions for geometry
-r seed                     random seed initialization
--rotation-angle [deg]      Angle of rotation
--rotation-steps n          Number of rotation steps
--coil ...                  configure type of coil
-h                          help


--pics--

Usage: pics [-l ...] [-r f] [-R ...] [-c] [-s f] [-i d] [-t <file>] [-n] [-N] [-g] [--gpu-gridding] [-p <file>] [--precond] [-b d] [-e] [-W <file>] [-d d] [-u f] [-C d] [-f f] [-I,--ist] [--fista] [-m,--admm] [-a,--pridu] [-w f] [-S] [--shared-img-dims d] [-K] [-B <file>] [-P f] [-M] [-U,--lowmem] [--no-toeplitz] [--psf_export <file>] [--psf_import <file>] [--wavelet <string>] [--mpi d] <kspace> <sensitivities> <output> 

Parallel-imaging compressed-sensing reconstruction.


-l 1/-l2                    toggle l1-wavelet or l2 regularization.
-r lambda                  regularization parameter
-R <T>:A:B:C               generalized regularization options (-Rh for help)
-c                         real-value constraint
-s step                    iteration stepsize
-i iter                    max. number of iterations
-t file                    k-space trajectory
-n                         disable random wavelet cycle spinning
-N                         do fully overlapping LLR blocks
-g                         use GPU
--gpu-gridding             use GPU for gridding
-p file                    pattern or weights
--precond                  interprete weights as preconditioner
-b blk                     Lowrank block size
-e                         Scale stepsize based on max. eigenvalue
-W <img>                   Warm start with <img>
-d level                   Debug level
-u rho                     ADMM rho
-C iter                    ADMM max. CG iterations
-f rfov                    restrict FOV
-I,--ist                   select IST
--fista                    select FISTA
-m,--admm                  select ADMM
-a,--pridu                 select Primal Dual
-w f                       inverse scaling of the data
-S                         re-scale the image after reconstruction
--shared-img-dims flags    deselect image dims with flags
-K                         randshift for NUFFT
-B file                    temporal (or other) basis
-P eps                     Basis Pursuit formulation, || y- Ax ||_2 <= eps
-M                         Simultaneous Multi-Slice reconstruction
-U,--lowmem                Use low-mem mode of the nuFFT
--no-toeplitz              Turn off Toeplitz mode of nuFFT
--psf_export file          Export PSF to file
--psf_import file          Import PSF from file
--wavelet name             wavelet type (haar,dau2,cdf44)
--mpi flags                distribute over this dimensions with use of MPI
-h                         help


--pocsense--

Usage: pocsense [-i d] [-r f] [-l d] <kspace> <sensitivities> <output> 

Perform POCSENSE reconstruction.

-i iter     max. number of iterations
-r alpha    regularization parameter
-l 1/-l2    toggle l1-wavelet or l2 regularization
-h          help


--poisson--

Usage: poisson [-Y d] [-Z d] [-y f] [-z f] [-C d] [-v] [-e] [-s d] <output> 

Computes Poisson-disc sampling pattern.

-Y size    size dimension 1
-Z size    size dimension 2
-y acc     acceleration dim 1
-z acc     acceleration dim 2
-C size    size of calibration region
-v         variable density
-e         elliptical scanning
-s seed    random seed
-h         help


--pol2mask--

Usage: pol2mask [-X d] [-Y d] <poly> <output> 

Compute masks from polygons.

-X size    size dimension 0
-Y size    size dimension 1
-h         help


--poly--

Usage: poly L N a_1 ... a_N <output> 

Evaluate polynomial p(x) = a_1 + a_2 x + a_3 x^2 ... a_(N+1) x^N at x = {0, 1, ... , L - 1} where a_i are floats.

-h  help


--psf--

Usage: psf <trajectory> <psf> 

Calculate point-spread-function (PSF) of given trajectory.


-h  help


--reconet--

Usage: reconet [-t,--train] [-e,--eval] [-a,--apply] [-g,--gpu] [-l,--load <file>] [-b,--batch-size d] [-I,--iterations d] [-n,--normalize] [-N,--network ...] [--resnet-block ...] [--varnet-block ...] [--tensorflow ...] [--unet ...] [--data-consistency ...] [--initial-reco ...] [--shared-weights] [--no-shared-weights] [--shared-lambda] [--no-shared-lambda] [--rss-norm] [--trajectory <file>] [--pattern <file>] [--mask <file>] [--valid-data ...] [--train-loss ...] [--valid-loss ...] [-T,--train-algo ...] [--adam ...] [--iPALM ...] [--load-memory] [--lowmem] [--test] [--export-graph <string>] <kspace> <sensitivities> <weights> <ref/out> 

Trains or appplies a neural network for reconstruction.

-t,--train                   train reconet
-e,--eval                    evaluate reconet
-a,--apply                   apply reconet
-g,--gpu                     run on gpu
-l,--load <weights-init>     load weights for continuing training
-b,--batch-size d            size of mini batches
-I,--iterations d            number of unrolled iterations
-n,--normalize               normalize data with maximum magnitude of adjoint reconstruction
-N,--network ...             select neural network
--resnet-block ...           configure residual block
--varnet-block ...           configure variational block
--tensorflow ...             configure tensorflow as network
--unet ...                   configure U-Net block
--data-consistency ...       configure data-consistency method
--initial-reco ...           configure initialization
--shared-weights             share weights across iterations
--no-shared-weights          share weights across iterations
--shared-lambda              share lambda across iterations
--no-shared-lambda           share lambda across iterations
--rss-norm                   scale output image to rss normalization
--trajectory <traj>          trajectory
--pattern <pattern>          sampling pattern / psf in kspace
--mask <mask>                mask for computation of loss
--valid-data ...             provide validation data
--train-loss ...             configure the training loss
--valid-loss ...             configure the validation loss
-T,--train-algo ...          configure general training parmeters
--adam ...                   configure Adam
--iPALM ...                  configure iPALM
--load-memory                copy training data into memory
--lowmem                     reduce memory usage by checkpointing
--test                       very small network for tests
--export-graph <file.dot>    export graph for visualization
-h                           help


--repmat--

Usage: repmat dimension repetitions <input> <output> 

Repeat input array multiple times along a certain dimension.

-h  help


--reshape--

Usage: reshape flags dim1 ... dimN <input> <output> 

Reshape selected dimensions.

-h  help


--resize--

Usage: resize [-c] dim1 size1 ... dimN sizeN <input> <output> 

Resizes an array along dimensions to sizes by truncating or zero-padding. Please see doc/resize.txt for examples.

-c    center
-h    help


--rmfreq--

Usage: rmfreq [-N d] [-M <file>] <traj> <k> <k_cor> 

Remove angle-dependent frequency

-N #       Number of harmonics [Default: 5]
-M file    Contrast modulation file
-h         help


--rof--

Usage: rof lambda flags <input> <output> 

Perform total variation denoising along dims <flags>.

-h  help


--roistat--

Usage: roistat [-b] [-C] [-S] [-M] [-D] [-E] [-V] <roi> <input> [<output>] 

Compute ROI statistics.

-b    Bessel's correction, i.e. 1 / (n - 1)
-C    voxel count
-S    sum
-M    mean
-D    standard deviation
-E    energy
-V    variance
-h    help


--rss--

Usage: rss bitmask <input> <output> 

Calculates root of sum of squares along selected dimensions.

-h  help


--rtnlinv--

Usage: rtnlinv [-i d] [-d d] [-c] [-N] [-m d] [-U] [-f f] [-p <file>] [-t <file>] [-I <file>] [-g] [-S] [-T f] [-x d:d:d] <kspace> <output> [<sensitivities>] 

Jointly estimate a time-series of images and sensitivities with nonlinear
inversion using {iter} iteration steps. Optionally outputs
the sensitivities.

-i iter         Number of Newton steps
-d level        Debug level
-c              Real-value constraint
-N              Do not normalize image with coil sensitivities
-m nmaps        Number of ENLIVE maps to use in reconstruction
-U              Do not combine ENLIVE maps in output
-f FOV          restrict FOV
-p file         pattern / transfer function
-t file         kspace trajectory
-I file         File for initialization
-g              use gpu
-S              Re-scale image after reconstruction
-T temp_damp    temporal damping [default: 0.9]
-x x:y:z        Explicitly specify image dimensions
-h              help


--sake--

Usage: sake [-i d] [-s f] <kspace> <output> 

Use SAKE algorithm to recover a full k-space from undersampled
data using low-rank matrix completion.

-i iter    number of iterations
-s size    rel. size of the signal subspace
-h         help


--saxpy--

Usage: saxpy scale <input1> <input2> <output> 

Multiply input1 with scale factor and add input2.

-h  help


--scale--

Usage: scale factor <input> <output> 

Scale array by {factor}. The scale factor can be a complex number.

-h  help


--sdot--

Usage: sdot <input1> <input2> 

Compute dot product along selected dimensions.

-h  help


--show--

Usage: show [-m] [-d d] [-s <string>] [-f <string>] <input> 

Outputs values or meta data.

-m           show meta data
-d dim       show size of dimension
-s sep       use <sep> as the separator
-f format    use <format> as the format. Default: "%%+.6e%%+.6ei"
-h           help


--signal--

Usage: signal [-F] [-B] [-T] [-S] [-M] [-G] [--fat] [-I] [-s] [-0 f:f:f] [-1 f:f:f] [-2 f:f:f] [-3 f:f:f] [-r f] [-e f] [-i f] [-f f] [-t f] [-n d] [-b d] [--av-spokes d] <basis-functions> 

Analytical simulation tool.

-F                FLASH
-B                bSSFP
-T                TSE
-S                SE
-M                MOLLI
-G                MGRE
--fat             Simulate additional fat component.
-I                inversion recovery
-s                inversion recovery starting from steady state
-0 min:max:N      range of off-resonance frequency (Hz)
-1 min:max:N      range of T1s (s)
-2 min:max:N      range of T2s (s)
-3 min:max:N      range of Mss
-r TR             repetition time
-e TE             echo time
-i TI             inversion time
-f FA             flip ange
-t T1 relax       T1 relax period (second) for MOLLI
-n n              number of measurements
-b heart beats    number of heart beats for MOLLI
--av-spokes d     Number of averaged consecutive spokes
-h                help


--sim--

Usage: sim [-1,--T1 f:f:f] [-2,--T2 f:f:f] [--BLOCH] [--BMC] [--ROT] [--ODE] [--STM] [--split-dim] [--seq ...] [--other ...] [--pool ...] <signal: Mxy> [<Partial derivatives: dR1, dM0, dR2, dB1>] 

simulation tool

-1,--T1 min:max:N    range of T1 values
-2,--T2 min:max:N    range of T2 values
--BLOCH              Bloch Equations (default)
--BMC                Bloch-McConnell Equations
--ROT                homogeneously discretized simulation based on rotational matrices
--ODE                full ordinary differential equation solver based simulation (default)
--STM                state-transition matrix based simulation
--split-dim          Split magnetization into x, y, and z component
--seq ...            configure sequence parameter
--other ...          configure other parameters
--pool ...           configure parameters for 2nd->5th pool
-h                   help


--slice--

Usage: slice dim1 pos1 ... dimN posN <input> <output> 

Extracts a slice from positions along dimensions.

-h  help


--spow--

Usage: spow exponent <input> <output> 

Raise array to the power of {exponent}. The exponent can be a complex number.

-h  help


--sqpics--

Usage: sqpics [-l ...] [-r f] [-R ...] [-s f] [-i d] [-t <file>] [-n] [-g] [-p <file>] [-b d] [-e] [-W <file>] [-d d] [-u f] [-C d] [-f f] [-m] [-w f] [-S] <kspace> <sensitivities> <output> 

Parallel-imaging compressed-sensing reconstruction.

-l 1/-l2         toggle l1-wavelet or l2 regularization.
-r lambda       regularization parameter
-R <T>:A:B:C    generalized regularization options (-Rh for help)
-s step         iteration stepsize
-i iter         max. number of iterations
-t file         k-space trajectory
-n              disable random wavelet cycle spinning
-g              use GPU
-p file         pattern or weights
-b blk          Lowrank block size
-e              Scale stepsize based on max. eigenvalue
-W <img>        Warm start with <img>
-d level        Debug level
-u rho          ADMM rho
-C iter         ADMM max. CG iterations
-f rfov         restrict FOV
-m              Select ADMM
-w val          scaling
-S              Re-scale the image after reconstruction
-h              help


--squeeze--

Usage: squeeze <input> <output> 

Remove singleton dimensions of array.

-h  help


--ssa--

Usage: ssa [-w d] [-z] [-m d] [-n d] [-r d] [-g d] <src> <EOF> [<S>] [<backprojection>] 

Perform SSA-FARY or Singular Spectrum Analysis. <src>: [samples, coordinates]

-w window     Window length
-z            Zeropadding [Default: True]
-m 0/1        Remove mean [Default: True]
-n 0/1        Normalize [Default: False]
-r rank       Rank for backprojection. r < 0: Throw away first r components. r > 0: Use only first r components.
-g bitmask    Bitmask for Grouping (long value!)
-h            help


--std--

Usage: std bitmask <input> <output> 

Compute standard deviation along selected dimensions specified by the {bitmask}

-h  help


--svd--

Usage: svd [-e] <input> <U> <S> <VH> 

Compute singular-value-decomposition (SVD).

-e    econ
-h    help


--tensorflow--

Usage: tensorflow [-b d] [-n] [-g] TensorFlow Graph [<Arguments>1> ... <Arguments>N> ] 

Load Tensorflow Graph

-b b    Fill placeholder in dims with b
-n      Print all nodes in graph
-g      Use gpu
-h      help


--tgv--

Usage: tgv lambda flags <input> <output> 

Perform total generalized variation denoising along dims specified by flags.

-h  help


--threshold--

Usage: threshold [-H] [-W] [-L] [-D] [-B] [-M] [-j d] [-b d] lambda <input> <output> 

Perform (soft) thresholding with parameter lambda.

-H              hard thresholding
-W              daubechies wavelet soft-thresholding
-L              locally low rank soft-thresholding
-D              divergence-free wavelet soft-thresholding
-B              thresholding with binary output where (val>lambda)
-M              thresholding with binary output where (val<lambda)
-j bitmask      joint soft-thresholding
-b blocksize    locally low rank block size
-h              help


--toimg--

Usage: toimg [-g f] [-c f] [-w f] [-d] [-m] [-W] [-D] <input> <output prefix> 

Create magnitude images as png or proto-dicom.
The first two non-singleton dimensions will
be used for the image, and the other dimensions
will be looped over.

-g gamma       gamma level
-c contrast    contrast level
-w window      window level
-d             write to dicom format (deprecated, use extension .dcm)
-m             re-scale each image
-W             use dynamic windowing
-D             Include dimensions in output filenames
-h             help


--traj--

Usage: traj [-x d] [-y d] [-d d] [-e d] [-a d] [-t d] [-m d] [-l] [-g] [-r] [-G] [-H] [-s d] [-A] [-D] [--double-base] [-o f] [-R f] [-q f:f:f] [-O] [-3] [-c] [-E] [-z d:d] [-C <file>] [--raga-inc d] <output> 

Computes k-space trajectories.

-x x             readout samples
-y y             phase encoding lines
-d d             full readout samples
-e e             number of echoes
-a a             acceleration
-t t             turns
-m mb            SMS multiband factor
-l               aligned partition angle
-g               golden angle in partition direction
-r               radial
-G               golden-ratio sampling
-H               halfCircle golden-ratio sampling
-s # Tiny GA     tiny golden angle
-A               rational approximation of golden angles
-D               projection angle in [0,360°), else in [0,180°)
--double-base    Define GA over 2Pi base instead of default Pi.
-o o             oversampling factor
-R phi           rotate
-q delays        gradient delays: x, y, xy
-O               correct transverse gradient error for radial tajectories
-3               3D
-c               asymmetric trajectory [DC sampled]
-E               multi-echo multi-spoke trajectory
-z Ref:Acel      Undersampling in z-direction.
-C file          custom_angle file [phi + i * psi]
--raga-inc d     Increment of RAGA Sampling
-h               help


--transpose--

Usage: transpose dim1 dim2 <input> <output> 

Transpose dimensions {dim1} and {dim2}.

-h  help


--twixread--

Usage: twixread [-x d] [-r d] [-y d] [-z d] [-s d] [-v d] [-c d] [-n d] [-p d] [-f d] [-i d] [-a d] [-A] [-L] [-P] [--rational] [-M] [--bin d] [-X] [-d d] <dat file> <output> 

Read data from Siemens twix (.dat) files.

-x X          number of samples (read-out)
-r R          radial lines
-y Y          phase encoding steps
-z Z          partition encoding steps
-s S          number of slices
-v V          number of averages
-c C          number of channels
-n N          number of repetitions
-p P          number of cardiac phases
-f F          number of flow encodings
-i I          number inversion experiments
-a A          total number of ADCs
-A            automatic [guess dimensions]
-L            use linectr offset
-P            use partctr offset
--rational    Rational Approximation Sampling
-M            MPI mode
--bin d       Binning of spokes for RAGA sampled data
-X            no consistency check for number of read acquisitions
-d level      Debug level
-h            help


--upat--

Usage: upat [-Y d] [-Z d] [-y d] [-z d] [-c d] <output> 

Create a sampling pattern.

-Y Y      size Y
-Z Z      size Z
-y uy     undersampling y
-z uz     undersampling z
-c cen    size of k-space center
-h        help


--var--

Usage: var bitmask <input> <output> 

Compute variance along selected dimensions specified by the {bitmask}

-h  help


--vec--

Usage: vec val1 ... valN <output> 

Create a vector of values.

-h  help


--version--

Usage: version [-t <string>] [-V] 

Print BART version. The version string is of the form
TAG or TAG-COMMITS-SHA as produced by 'git describe'. It
specifies the last release (TAG), and (if git is used)
the number of commits (COMMITS) since this release and
the abbreviated hash of the last commit (SHA). If there
are local changes '-dirty' is added at the end.

-t version    Check minimum version
-V            Output verbose info
-h            help


--walsh--

Usage: walsh [-r d:d:d] [-b d:d:d] <input> <output> 

Estimate coil sensitivities using walsh method (use with ecaltwo).

-r cal_size      Limits the size of the calibration region.
-b block_size    Block size.
-h               help


--wave--

Usage: wave [-r f] [-b d] [-i d] [-s f] [-c f] [-t f] [-e f] [-g] [-f] [-H] [-v] [-w] [-l] <maps> <wave> <kspace> <output> 

Perform a wave-caipi 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).
Expected dimensions:
  * maps    - ( sx, sy, sz, nc, md)
  * wave    - ( wx, sy, sz,  1,  1)
  * kspace  - ( wx, sy, sz, nc,  1)
  * output  - ( sx, sy, sz,  1, md)

-r lambda    Soft threshold lambda for wavelet or locally low rank.
-b blkdim    Block size for locally low rank.
-i mxiter    Maximum number of iterations.
-s stepsz    Step size for iterative method.
-c cntnu     Continuation value for IST/FISTA.
-t toler     Tolerance convergence condition for iterative method.
-e eigvl     Maximum eigenvalue of normal operator, if known.
-g           use GPU
-f           Reconstruct using FISTA instead of IST.
-H           Use hogwild in IST/FISTA.
-v           Split result to real and imaginary components.
-w           Use wavelet.
-l           Use locally low rank across the real and imaginary components.
-h           help


--wavelet--

Usage: wavelet [-a] [-H] [-D] [-C] bitmask [dim1 ... dimN ] <input> <output> 

Perform wavelet transform.

-a    adjoint (specify dims)
-H    type: Haar
-D    type: Dau2
-C    type: CDF44
-h    help


--wavepsf--

Usage: wavepsf [-c] [-x d] [-y d] [-r f] [-a d] [-t f] [-g f] [-s f] [-n d] <output> 

Generate a wave PSF in hybrid space.
- Assumes the first dimension is the readout dimension.
- Only generates a 2 dimensional PSF.
- Use reshape and fmac to generate a 3D PSF.

3D PSF Example:
bart wavepsf		-x 768 -y 128 -r 0.1 -a 3000 -t 0.00001 -g 0.8 -s 17000 -n 6 wY
bart wavepsf -c -x 768 -y 128 -r 0.1 -a 3000 -t 0.00001 -g 0.8 -s 17000 -n 6 wZ
bart reshape 7 wZ 768 1 128 wZ wZ
bart fmac wY wZ wYZ

-c           Set to use a cosine gradient wave
-x RO_dim    Number of readout points
-y PE_dim    Number of phase encode points
-r PE_res    Resolution of phase encode in cm
-a ADC_T     Readout duration in microseconds.
-t ADC_dt    ADC sampling rate in seconds
-g gMax      Maximum gradient amplitude in Gauss/cm
-s sMax      Maximum gradient slew rate in Gauss/cm/second
-n ncyc      Number of cycles in the gradient wave
-h           help


--whiten--

Usage: whiten [-o <file>] [-c <file>] [-n] <input> <ndata> <output> [<optmat_out>] [<covar_out>] 

Apply multi-channel noise pre-whitening on <input> using noise data <ndata>.
Optionally output whitening matrix and noise covariance matrix

-o <optmat_in>    use external whitening matrix <optmat_in>
-c <covar_in>     use external noise covariance matrix <covar_in>
-n                normalize variance to 1 using noise data <ndata>
-h                help


--window--

Usage: window [-H] flags <input> <output> 

Apply Hamming (Hann) window to <input> along dimensions specified by flags

-H    Hann window
-h    help


--wshfl--

Usage: wshfl [-R ...] [-b d] [-i d] [-j d] [-s f] [-e f] [-F <file>] [-O <file>] [-t f] [-g] [-K] [-H] [-v] <maps> <wave> <phi> <reorder> <table> <output> 

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)

-R <T>:A:B:C    Generalized regularization options. (-Rh for help)
-b blkdim       Block size for locally low rank.
-i mxiter       Maximum number of iterations.
-j cgiter       Maximum number of CG iterations in ADMM.
-s admrho       ADMM Rho value.
-e eigval       Eigenvalue to scale step size. (Optional.)
-F frwrd        Go from shfl-coeffs to data-table. Pass in coeffs path.
-O initl        Initialize reconstruction with guess.
-t toler        Tolerance convergence condition for FISTA.
-g              Use GPU.
-K              Go from data-table to shuffling basis k-space.
-H              Use hogwild.
-v              Split coefficients to real and imaginary components.
-h              help


--zeros--

Usage: zeros dims dim1 ... dimN <output> 

Create a zero-filled array with {dims} dimensions of size {dim1} to {dimn}.

-h  help


--zexp--

Usage: zexp [-i] <input> <output> 

Point-wise complex exponential.

-i    imaginary
-h    help
