Edit on GitHub

Acquisition Data

There are two different forms of acquisition data types in MRIReco:

  • RawAcquisitionData
  • AcquisitionData

While the former is used to hold the data in the form, how it will be written out from the scanner, the later has already performed some data permutations bringing the data into the shape how the reconstruction expects it.

Raw Data

The RawAcquisitionData is a data type that closely resembles the ISMRMRD data format. It looks like

mutable struct RawAcquisitionData
  params::Dict{String, Any}
  profiles::Vector{Profile}
end

with

mutable struct Profile
  head::AcquisitionHeader
  traj::Array{Float32,2}
  data::Array{Complex{Float32},2}
end

The params member of RawAcquisitionData is basically a flattened dictionary derived from the XML part of an ISMRMRD file. A Profile describes the data measured after a single excitation during an MRI experiment. It has members head, traj, and data, which exactly resemble the structures specified by the ISMRMRD file format.

AcquisitionHeader has exactly the same structure as ISMRMRD. You can find more information about it here

Two fields are especially important in it :

  • flags
  • idx

flags

The flags field in the AcquisitionHeader is a 64 bit mask that can be used to indicate specific attributes of the corresponding readout. One usage of these flags is to reverse the signal during conversion from RawAcquisitionData to AcquisitionData if the flag "ACQISREVERSE" is set.

FLAGS = Dict(
    "ACQ_FIRST_IN_ENCODE_STEP1"                => 1,
    "ACQ_LAST_IN_ENCODE_STEP1"                 => 2,
    "ACQ_FIRST_IN_ENCODE_STEP2"                => 3,
    "ACQ_LAST_IN_ENCODE_STEP2"                 => 4,
    "ACQ_FIRST_IN_AVERAGE"                     => 5,
    "ACQ_LAST_IN_AVERAGE"                      => 6,
    "ACQ_FIRST_IN_SLICE"                       => 7,
    "ACQ_LAST_IN_SLICE"                        => 8,
    "ACQ_FIRST_IN_CONTRAST"                    => 9,
    "ACQ_LAST_IN_CONTRAST"                     => 10,
    "ACQ_FIRST_IN_PHASE"                       => 11,
    "ACQ_LAST_IN_PHASE"                        => 12,
    "ACQ_FIRST_IN_REPETITION"                  => 13,
    "ACQ_LAST_IN_REPETITION"                   => 14,
    "ACQ_FIRST_IN_SET"                         => 15,
    "ACQ_LAST_IN_SET"                          => 16,
    "ACQ_FIRST_IN_SEGMENT"                     => 17,
    "ACQ_LAST_IN_SEGMENT"                      => 18,
    "ACQ_IS_NOISE_MEASUREMENT"                 => 19,
    "ACQ_IS_PARALLEL_CALIBRATION"              => 20,
    "ACQ_IS_PARALLEL_CALIBRATION_AND_IMAGING"  => 21,
    "ACQ_IS_REVERSE"                           => 22,
    "ACQ_IS_NAVIGATION_DATA"                   => 23,
    "ACQ_IS_PHASECORR_DATA"                    => 24,
    "ACQ_LAST_IN_MEASUREMENT"                  => 25,
    "ACQ_IS_HPFEEDBACK_DATA"                   => 26,
    "ACQ_IS_DUMMYSCAN_DATA"                    => 27,
    "ACQ_IS_RTFEEDBACK_DATA"                   => 28,
    "ACQ_IS_SURFACECOILCORRECTIONSCAN_DATA"    => 29,
    "ACQ_COMPRESSION1"                         => 53,
    "ACQ_COMPRESSION2"                         => 54,
    "ACQ_COMPRESSION3"                         => 55,
    "ACQ_COMPRESSION4"                         => 56,
    "ACQ_USER1"                                => 57,
    "ACQ_USER2"                                => 58,
    "ACQ_USER3"                                => 59,
    "ACQ_USER4"                                => 60,
    "ACQ_USER5"                                => 61,
    "ACQ_USER6"                                => 62,
    "ACQ_USER7"                                => 63,
    "ACQ_USER8"                                => 64
)

You can check the flags of a profile with flags_of(p:Profile OR head::AcquisitionHeader) or flag_is_set and manipulate them with thus functions :

  • flag_set!(obj::Profile, flag)
  • flag_remove!(obj::Profile, flag)
  • flag_remove_all!(obj::Profile)

Thus functions can also been directly applied to the AcquisitionHeader and you can set! or remove! multiple flags at once :

flag_set!(head::AcquisitionHeader, [FLAGS["ACQ_USER8"], FLAGS["ACQ_IS_REVERSE"]])
# OR
flag_set!(head::AcquisitionHeader, [FLAGS["ACQ_USER8"],FLAGS["ACQ_IS_REVERSE"]])

Alternatively, you can set the flags directly with specific Const variable :

h = AcquisitionHeader()
h.flags = ACQ_IS_PARALLEL_CALIBRATION | ACQ_IS_NOISE_MEASUREMENT
flags_of(h)
h.flags = h.flags & ~ACQ_IS_NOISE_MEASUREMENT # remove the flag ACQ_IS_NOISE_MEASUREMENT
flags_of(h)

idx

MR acquisitions often loop through a set of counters (e.g. phase encodes) in a complete experiment. The following encoding counters are referred to by the idx field in the AcquisitionHeader (See the ISMRMRD documentation)

Preprocessed Data

The RawAcquisitionData can be preprocessed into a form, which makes it more convenient for reconstruction algorithms. The AcquisitionData type looks like

mutable struct AcquisitionData
  sequenceInfo::Dict{Symbol,Any}
  traj::Vector{Trajectory}
  kdata::Array{Matrix{ComplexF64},3}
  subsampleIndices::Vector{Array{Int64}}
  encodingSize::Vector{Int64}
  fov::Vector{Float64}
end

It consists of the sequence informations stored in a dictionary, the k-space trajectory, the k-space data, and several parameters describing the dimension of the data and some additional index vectors.

The k-space data kdata has three dimensions encoding

  1. dim : contrasts/echoes
  2. dim : slices
  3. dim : repetitions

Each element is a matrix encoding

  1. dim : k-space nodes
  2. dim : channels/coils

In case of undersampled data, the subsampling indices are stored in subsampleIndices. One check if the data is undersampled by checking if isempty(subsampleIndices).

The encoded space is stored in the field encodingSize. It is especially relevant for non-Cartesian trajectories where it is not clear upfront, how large the grid size for reconstruction should be chosen. Finally fov describes the physical lengths of the encoding grid.

Manipulating rawData

Extract a subset of profiles with flags

A subset of rawData can be extracted according to the flags using the function filter_raw_by_flags. For example this function can be used to extract :

  • Noise : raw_noise = filter_raw_by_flags(raw,"ACQ_IS_NOISE_MEASUREMENT")
  • Extract calibration data : raw_noise = filter_raw_by_flags(raw,["ACQ_IS_PARALLEL_CALIBRATION_AND_IMAGING","ACQ_IS_PARALLEL_CALIBRATION"])

If you want to retrieve the data without the noise and calibration data you can use the following command :

raw = remove_raw_by_flags(raw,["ACQ_IS_NOISE_MEASUREMENT","ACQ_IS_PARALLEL_CALIBRATION_AND_IMAGING","ACQ_IS_PARALLEL_CALIBRATION"])

by default the following flags are removed :

"ACQ_IS_NOISE_MEASUREMENT", 
"ACQ_IS_PARALLEL_CALIBRATION", 
"ACQ_IS_NAVIGATION_DATA", 
"ACQ_IS_PHASECORR_DATA", 
"ACQ_IS_DUMMYSCAN_DATA", 
"ACQ_IS_PHASE_STABILIZATION_REFERENCE", "ACQ_IS_PHASE_STABILIZATION"

Pre-processing tools

Remove readout oversampling

On clinical MR scanners, the raw data are often oversampled in the readout direction. This oversampling is not necessary for image reconstruction and can be removed to save memory and computation time. This step is generally applied right after the acquisition of each line. The current implementation is performed on the RawAcquisitionData (before conversion to AcquisitionData) and is based on the following parameters:

  • rawData.params["encodedFOV"][1]
  • rawData.params["reconFOV"][1]

If an oversampling factor of 2 is applied during acquisition, the encodedFOV will be 2 times larger than reconFOV along the readout dimension.

!!! note
    Be careful to remove all the profiles from a `RawAcquisitionData` that are not supposed to be of the same size; like noise acquisition or navigators. You can use the function `raw = remove_raw_by_flags(raw)`.