Skip to content

Quantitative MRI

Quantitative MRI (qMRI) is a collection of methods aiming at generating parametric maps that can characterize underlying tissue properties. Unlike those of conventional MR images (for example, T1w or T2w), intensity values of quantitative maps are not represented in an arbitrary range. Instead, these maps are represented either in absolute physical units (for example, seconds for T1map), or within an application dependent range of arbitrary units (for example, myelin water fraction MWFmap in brain).

Organization of qMRI data in BIDS

Unlike conventional MR images, quantitative maps are not immediate products of the image reconstruction step (from k-space data to structural images). Intensity values of qMRI maps are calculated by fitting a collection of parametrically linked images to a biophysical model or to an MRI signal representation. This processing is typically carried out in the image domain. There are two main ways to obtain a quantitative map:

  1. Pre-generated qMRI maps: The qMRI maps are generated right after the reconstruction of required input images and made available to the user at the scanner console. The acquisition scenarios may include (a) vendor pipelines or (b) open-source pipelines deployed at the scanner site.

  2. Post-generated qMRI maps: The qMRI maps are generated from a collection of input data after they are exported from the scanner site. This type of processing is commonly carried out using an open-source software such as hMRI toolbox, mrQ, PyQMRI, qmap, qMRLab, and QUIT.

Inputs are file collections

The common concept of entity-linked file collections enables the description of a qMRI application by creating logical groups of input files through suffix and certain entities representing acquisition parameters (echo, flip, inv, mt) or file parts (part).

If a qMRI file collection is intended for creating structural quantitative maps (for example, T1map), files belonging to that collection are stored in the anat subdirectory.

List of currently supported collections:

Name suffix Description
Inversion recovery T1 mapping IRT1 The IRT1 method involves multiple inversion recovery spin-echo images acquired at different inversion times (Barral et al. 2010).
Multi-echo Gradient Recalled Echo MEGRE Anatomical gradient echo images acquired at different echo times. Please note that this suffix is not intended for the logical grouping of images acquired using an Echo Planar Imaging (EPI) readout.
Multi-echo Spin Echo MESE The MESE method involves multiple spin echo images acquired at different echo times and is primarily used for T2 mapping. Please note that this suffix is not intended for the logical grouping of images acquired using an Echo Planar Imaging (EPI) readout.
Magnetization Prepared Two Gradient Echoes MP2RAGE The MP2RAGE method is a special protocol that collects several images at different flip angles and inversion times to create a parametric T1map by combining the magnitude and phase images (Marques et al. 2010).
Multi-parametric Mapping MPM The MPM approaches (a.k.a hMRI) involves the acquisition of highly-similar anatomical images that differ in terms of application of a magnetization transfer RF pulse (MTon or MToff), flip angle and (optionally) echo time and magnitue/phase parts (Weiskopf et al. 2013). See here for suggested MPM acquisition protocols.
Magnetization Transfer Ratio MTR This method is to calculate a semi-quantitative magnetization transfer ratio map.
Magnetization transfer saturation MTS This method is to calculate a semi-quantitative magnetization transfer saturation index map. The MTS method involves three sets of anatomical images that differ in terms of application of a magnetization transfer RF pulse (MTon or MToff) and flip angle (Helms et al. 2008).
Variable flip angle VFA The VFA method involves at least two spoiled gradient echo (SPGR) of steady-state free precession (SSFP) images acquired at different flip angles. Depending on the provided metadata fields and the sequence type, data may be eligible for DESPOT1, DESPOT2 and their variants (Deoni et al. 2005).

Template:

sub-<label>/
    [ses-<label>/]
        anat/
            sub-<label>[_ses-<label>][_task-<label>][_acq-<label>][_ce-<label>][_rec-<label>][_run-<index>]_echo-<index>[_part-<mag|phase|real|imag>][_chunk-<index>]_MEGRE.json
            sub-<label>[_ses-<label>][_task-<label>][_acq-<label>][_ce-<label>][_rec-<label>][_run-<index>]_echo-<index>[_part-<mag|phase|real|imag>][_chunk-<index>]_MEGRE.nii[.gz]
            sub-<label>[_ses-<label>][_task-<label>][_acq-<label>][_ce-<label>][_rec-<label>][_dir-<label>][_run-<index>]_echo-<index>[_part-<mag|phase|real|imag>][_chunk-<index>]_MESE.json
            sub-<label>[_ses-<label>][_task-<label>][_acq-<label>][_ce-<label>][_rec-<label>][_dir-<label>][_run-<index>]_echo-<index>[_part-<mag|phase|real|imag>][_chunk-<index>]_MESE.nii[.gz]
            sub-<label>[_ses-<label>][_task-<label>][_acq-<label>][_ce-<label>][_rec-<label>][_run-<index>][_echo-<index>]_flip-<index>[_part-<mag|phase|real|imag>][_chunk-<index>]_VFA.json
            sub-<label>[_ses-<label>][_task-<label>][_acq-<label>][_ce-<label>][_rec-<label>][_run-<index>][_echo-<index>]_flip-<index>[_part-<mag|phase|real|imag>][_chunk-<index>]_VFA.nii[.gz]
            sub-<label>[_ses-<label>][_task-<label>][_acq-<label>][_ce-<label>][_rec-<label>][_run-<index>]_inv-<index>[_part-<mag|phase|real|imag>][_chunk-<index>]_IRT1.json
            sub-<label>[_ses-<label>][_task-<label>][_acq-<label>][_ce-<label>][_rec-<label>][_run-<index>]_inv-<index>[_part-<mag|phase|real|imag>][_chunk-<index>]_IRT1.nii[.gz]
            sub-<label>[_ses-<label>][_task-<label>][_acq-<label>][_ce-<label>][_rec-<label>][_run-<index>][_echo-<index>][_flip-<index>]_inv-<index>[_part-<mag|phase|real|imag>][_chunk-<index>]_MP2RAGE.json
            sub-<label>[_ses-<label>][_task-<label>][_acq-<label>][_ce-<label>][_rec-<label>][_run-<index>][_echo-<index>][_flip-<index>]_inv-<index>[_part-<mag|phase|real|imag>][_chunk-<index>]_MP2RAGE.nii[.gz]
            sub-<label>[_ses-<label>][_task-<label>][_acq-<label>][_ce-<label>][_rec-<label>][_run-<index>][_echo-<index>]_flip-<index>_mt-<on|off>[_part-<mag|phase|real|imag>][_chunk-<index>]_MPM.json
            sub-<label>[_ses-<label>][_task-<label>][_acq-<label>][_ce-<label>][_rec-<label>][_run-<index>][_echo-<index>]_flip-<index>_mt-<on|off>[_part-<mag|phase|real|imag>][_chunk-<index>]_MPM.nii[.gz]
            sub-<label>[_ses-<label>][_task-<label>][_acq-<label>][_ce-<label>][_rec-<label>][_run-<index>][_echo-<index>]_flip-<index>_mt-<on|off>[_part-<mag|phase|real|imag>][_chunk-<index>]_MTS.json
            sub-<label>[_ses-<label>][_task-<label>][_acq-<label>][_ce-<label>][_rec-<label>][_run-<index>][_echo-<index>]_flip-<index>_mt-<on|off>[_part-<mag|phase|real|imag>][_chunk-<index>]_MTS.nii[.gz]
            sub-<label>[_ses-<label>][_task-<label>][_acq-<label>][_ce-<label>][_rec-<label>][_run-<index>]_mt-<on|off>[_part-<mag|phase|real|imag>][_chunk-<index>]_MTR.json
            sub-<label>[_ses-<label>][_task-<label>][_acq-<label>][_ce-<label>][_rec-<label>][_run-<index>]_mt-<on|off>[_part-<mag|phase|real|imag>][_chunk-<index>]_MTR.nii[.gz]
Legend:

  • For more information about filename elements (for example, entities, suffixes, extensions), follow the links embedded in the filename template.

  • <matches> is a placeholder to denote an arbitrary (and valid) sequence of entities and labels at the beginning of the filename (only BIDS "raw").

  • <source-entities> is a placeholder to denote an arbitrary sequence of entities and labels at the beginning of the filename matching a source file from which the file derives (only BIDS-Derivatives).

  • Filename entities or directories between square brackets (for example, [_ses-<label>]) are OPTIONAL.

  • Some entities may only allow specific values, in which case those values are listed in <>, separated by |.

  • _<suffix> means that there are several (>6) valid suffixes for this filename pattern.

  • .<extension> means that there are several (>6) valid extensions for this file type.

  • [.gz] means that both the unzipped and gzipped versions of the extension are valid.

Below is an example file collection for MP2RAGE:

└─ sub-01/
   └─ anat/
      ├─ sub-01_inv-1_part-mag_MP2RAGE.nii.gz 
      ├─ sub-01_inv-1_part-phase_MP2RAGE.nii.gz 
      ├─ sub-01_inv-1_MP2RAGE.json 
      ├─ sub-01_inv-2_part-mag_MP2RAGE.nii.gz 
      ├─ sub-01_inv-2_part-phase_MP2RAGE.nii.gz 
      └─ sub-01_inv-2_MP2RAGE.json

Commonly, RF fieldmaps (B1+ and B1- maps) are used for the correction of structural quantitative maps. As these images do not convey substantial structural information, respective file collections of RF fieldmaps are stored in the fmap subdirectory. Below is an example file collection for RF transmit field map TB1EPI:

└─ sub-01/
   └─ fmap/
      ├─ sub-01_echo-1_flip-1_TB1EPI.nii.gz 
      ├─ sub-01_echo-1_flip-1_TB1EPI.json 
      ├─ sub-01_echo-2_flip-1_TB1EPI.nii.gz 
      ├─ sub-01_echo-2_flip-1_TB1EPI.json 
      ├─ sub-01_echo-1_flip-2_TB1EPI.nii.gz 
      ├─ sub-01_echo-1_flip-2_TB1EPI.json 
      ├─ sub-01_echo-2_flip-2_TB1EPI.nii.gz 
      └─ sub-01_echo-2_flip-2_TB1EPI.json

Please visit the file collections appendix to see the list of currently supported qMRI applications.

Outputs are quantitative maps

qMRI maps are stored differently depending on the process that generated them. Pre-generated qMRI maps MAY be stored as part of a raw BIDS dataset, whereas they MUST be stored in a derivative BIDS dataset if they were post-generated.

See the example below of a T1map generated from an MP2RAGE file collection using either option.

If the map is post-generated:

└─ ds-example/
   ├─ derivatives/
   │  └─ qMRI-software-name/
   │     └─ sub-01/
   │        └─ anat/
   │           ├─ sub-01_T1map.nii.gz  # --> T1 map in a derivative dataset
   │           ├─ sub-01_T1map.json 
   │           ├─ sub-01_UNIT1.nii.gz  # --> UNI T1 in a derivative dataset
   │           └─ sub-01_UNIT1.json 
   └─ sub-01/
      └─ anat/
         ├─ sub-01_inv-1_part-mag_MP2RAGE.nii.gz 
         ├─ sub-01_inv-1_part-phase_MP2RAGE.nii.gz 
         ├─ sub-01_inv-1_MP2RAGE.json 
         ├─ sub-01_inv-2_part-mag_MP2RAGE.nii.gz 
         ├─ sub-01_inv-2_part-phase_MP2RAGE.nii.gz 
         └─ sub-01_inv-2_MP2RAGE.json

If the map is pre-generated, for example, by a Siemens scanner:

└─ ds-example/
   └─ sub-01/
      └─ anat/
         ├─ sub-01_inv-1_part-mag_MP2RAGE.nii.gz 
         ├─ sub-01_inv-1_part-phase_MP2RAGE.nii.gz 
         ├─ sub-01_inv-1_MP2RAGE.json 
         ├─ sub-01_inv-2_part-mag_MP2RAGE.nii.gz 
         ├─ sub-01_inv-2_part-phase_MP2RAGE.nii.gz 
         ├─ sub-01_inv-2_MP2RAGE.json 
         ├─ sub-01_T1map.nii.gz  # --> T1 map in a raw dataset
         ├─ sub-01_T1map.json 
         ├─ sub-01_UNIT1.nii.gz  # --> UNI T1 in a raw dataset
         └─ sub-01_UNIT1.json

Sharing of vendor outputs

Even though the process from which pre-generated qMRI maps are obtained (vendor pipelines) is not known, vendors generally allow exporting of the corresponding input data. It is RECOMMENDED to share them along with the vendor outputs, whenever possible for a qMRI method supported by BIDS.

Example datasets

You can find example file collections and qMRI maps organized according to BIDS in the BIDS examples.

Metadata requirements for qMRI data

The table of required entities for qMRI file collections are provided in the entity table. However, viability of a qMRI file collection is determined not only by the naming and organization of the input files, but also by which metadata fields are provided in accompanying json files.

Method-specific priority levels for qMRI file collections

Anatomy imaging data

File collection REQUIRED metadata OPTIONAL metadata
VFA FlipAngle, PulseSequenceType, RepetitionTimeExcitation SpoilingRFPhaseIncrement
IRT1 InversionTime
MP2RAGE* FlipAngle, InversionTime, RepetitionTimeExcitation, RepetitionTimePreparation, NumberShots,MagneticFieldStrength EchoTime
MESE EchoTime
MEGRE EchoTime
MTR MTState
MTS FlipAngle, MTState, RepetitionTimeExcitation
MPM FlipAngle, MTState, RepetitionTimeExcitation EchoTime

* Please see MP2RAGE-specific notes for the calculation of NumberShots and regarding the organization of UNIT1 image.

Explanation of the table:

  • The metadata fields listed in the REQUIRED column are needed to perform a minimum viable qMRI processing for the corresponding file collection.

  • Note that some of the metadata fields may be constant across different files in a file collection, yet still required as an input (for example, NumberShots in MP2RAGE). Such metadata fields MUST be provided in the accompanying JSON files.

  • The metadata fields listed in the OPTIONAL column can be used to form different flavors of an existing file collection suffix, dispensing with the need for introducing a new suffix. See deriving the intended qMRI application from an ambiguous file collection for details.

Metadata requirements for qMRI maps

As qMRI maps are stored as derivatives, they are subjected to the metadata requirements of derived datasets.

An example dataset_description.json for a qMRI map derivatives directory:

└─ ds-example/
   └─ derivatives/
      └─ qMRLab/
         ├─ dataset_description.json 
         └─ sub-01/
            └─ anat/
               ├─ sub-01_T1map.nii.gz 
               ├─ sub-01_T1map.json 
               ├─ sub-01_M0map.nii.gz 
               └─ sub-01_M0map.json

dataset_description.json:

{
  "Name": "qMRLab Outputs",
  "BIDSVersion": "1.5.0",
  "DatasetType": "derivative",
  "GeneratedBy": [
    {
      "Name": "qMRLab",
      "Version": "2.4.1",
      "Container": {
        "Type": "docker",
        "Tag": "qmrlab/minimal:2.4.1"
        }
    },
    {
      "Name": "Manual",
      "Description": "Generated example T1map outputs"
    }
  ],
  "SourceDatasets": [
    {
      "DOI": "doi:10.17605/OSF.IO/K4BS5",
      "URL": "https://osf.io/k4bs5/",
      "Version": "1"
    }
  ]
}

In addition to the metadata fields provided in the dataset_description.json, qMRI maps are RECOMMENDED to be accompanied by sidecar JSON files that contain further information about the quantified maps. Although this may not be the generic case for common derivative outputs, a proper interpretation of qMRI maps may critically depend on some metadata fields. For example, without the information of MagneticFieldStrength, white-matter T1 values in a T1map become elusive.

  • All the acquisition parameters that are constant across the files in a file collection are RECOMMENDED to be added to the sidecar json of the qMRI maps.

  • Relevant acquisition parameters that vary across files in a qMRI file collection are RECOMMENDED to be added to the sidecar json of the qMRI map in array form.

  • The JSON file accompanying a qMRI map which is obtained by using open-source software is RECOMMENDED to include additional metadata fields listed in the following table:

Key name Requirement Level Data type Description
Sources RECOMMENDED array of strings A list of files with the paths specified using BIDS URIs; these files were directly used in the creation of this derivative data file. For example, if a derivative A is used in the creation of another derivative B, which is in turn used to generate C in a chain of A->B->C, C should only list B in "Sources", and B should only list A in "Sources". However, in case both X and Y are directly used in the creation of Z, then Z should list X and Y in "Sources", regardless of whether X was used to generate Y. Using paths specified relative to the dataset root is DEPRECATED.
EstimationReference RECOMMENDED string Reference to the study/studies on which the implementation is based.
EstimationAlgorithm RECOMMENDED string Type of algorithm used to perform fitting (for example, "linear", "non-linear", "LM" and such).
Units RECOMMENDED string Measurement units for the associated variable. SI units in CMIXF formatting are RECOMMENDED (see Units).
BasedOn DEPRECATED string or array of strings List of files in a file collection to generate the map. Fieldmaps are also listed, if involved in the processing. This field is DEPRECATED, and this metadata SHOULD be recorded in the Sources field using BIDS URIs to distinguish sources from different datasets.

Example:

sub-01_T1map.nii.gz
sub-01_T1map.json

sub-01_T1map.json:

{

<<Parameter injected by the software/pipeline>>

"Sources":["bids:raw:sub-01/anat/sub-01_flip-1_VFA.nii.gz",
           "bids:raw:sub-01/anat/sub-01_flip-2_VFA.nii.gz",
           "bids:raw:sub-01/anat/sub-01_flip-3_VFA.nii.gz",
           "bids:raw:sub-01/anat/sub-01_flip-4_VFA.nii.gz",
           "bids:raw:sub-01/fmap/sub-01_TB1map.nii.gz"],
"EstimationPaper":"Deoni et. al.MRM, 2015",
"EstimationAlgorithm":"Linear",
"Units": "second",

<<Parameters that are constant across files in the (parent) file collection>>

"MagneticFieldStrength": "3",
"Manufacturer": "Siemens",
"ManufacturerModelName": "TrioTim",
"InstitutionName": "xxx",
"PulseSequenceType": "SPGR",
"PulseSequenceDetails": "Information beyond the sequence type that identifies
 specific pulse sequence used (VB version, if not standard, Siemens WIP XXX
 ersion ### sequence written by xx using a version compiled on mm/dd/yyyy/)",
"RepetitionTimeExcitation": "35",
"EchoTime": "2.86",
"SliceThickness": "5",

<<Relevant parameters that vary across the linking entity of the (parent) file collection>>

"FlipAngle": ["5","10","15","20"]

}

Deriving the intended qMRI application from an ambiguous file collection

Certain file collection suffixes may refer to a generic data collection regime such as variable flip angle (VFA), rather than a more specific acquisition, for example, magnetization prepared two gradient echoes (MP2RAGE). Such generic acquisitions can serve as a basis to derive various qMRI applications by changes to the acquisition sequence (for example, readout) type or by varying additional scan parameters.

If such an inheritance relationship is applicable between an already existing file collection and a new qMRI application to be included in the specification, the inheritor qMRI method is listed in the table below instead of introducing a new file collection suffix. This approach aims at:

  • preventing the list of available suffixes from over-proliferation,
  • providing qMRI-focused BIDS applications with a set of meta-data driven rules to infer possible fitting options,
  • keeping an inheritance track of the qMRI methods described within the specification.
File-collection suffix If REQUIRED metadata == Value OPTIONAL metadata (entity/fixed) Derived application name (NOT a suffix)
VFA PulseSequenceType == SPGR DESPOT1
VFA PulseSequenceType == SSFP SpoilingRFPhaseIncrement (fixed) DESPOT2
MP2RAGE EchoTime (echo) MP2RAGE-ME
MPM EchoTime (echo) MPM-ME

In this table, (entity/fixed) denotes whether the OPTIONAL metadata that forms a new flavor of qMRI application for the respective suffix varies across files of a file collection (which calls for using a linking entity) or fixed. If former is the case, the entity is to be added to the files in that file collection. Note that this addition MUST be allowed by the priority levels given for that suffix in the entity table. If latter (fixed) is the case, filenames will remain the same; however, the optional metadata (third column) may define the flavor of the application (fourth column) along with the conditional value of a required metadata field (second column).

A derived qMRI application becomes available if all the optional metadata fields listed for the respective file collection suffix are provided for the data. In addition, conditional rules based on the value of a given required metadata field can be set for the description of a derived qMRI application. Note that the value of this required metadata is fixed across constituent images of a file collection and defined in Method-specific priority levels for qMRI file collections.

For example, if the optional metadata field of PulseSequenceType is SPGR for a collection of anatomical images listed by the VFA suffix, the data qualifies for DESPOT1 T1 fitting. For the same suffix, if the PulseSequenceType metadata field has the value of SSFP, and the SpoilingRFPhaseIncrement is provided as a metadata field, then the dataset becomes eligible for DESPOT2 T2 fitting application.

Please note that optional metadata fields listed in the deriving the intended qMRI application from an ambiguous file collection table are included in the optional (third) column of the priority levels table for the consistency of this appendix.

Introducing a new qMRI file collection

If a qMRI application cannot be interpreted as a subtype of an already existing suffix of a qMRI-related file collection, we RECOMMEND adhering to the following principles to introduce a new suffix:

  • All qMRI-relevant file collection suffixes are capitalized.

  • Unless the pulse sequence is exclusively associated with a specific qMRI application (for example, MP2RAGE), sequence names are not used as suffixes.

  • File collection suffixes for qMRI applications attain a clear description of the qMRI method that they relate to in the file collections appendix.

  • Hyperlinks to example applications and reference method articles are encouraged whenever possible.

  • If it is possible to derive a qMRI application from an already existing file collection suffix by defining a set of logical conditions over the metadata fields, the tables of the deriving the intended qMRI application from an ambiguous file collection and the anatomy data priority levels sections are extended instead of introducing a new suffix.

Application-specific notes for qMRI file collections

Anatomy imaging data

General notes:

  • Some BIDS metadata field values are calculated based on the values of other metadata fields that are not listed as required fields. These fields include: NumberShots. The calculation of the values may depend on the type of the acquisition. These acquisitions include: MP2RAGE and TB1SRGE.

MP2RAGE specific notes

UNIT1 images

Although the UNIT1 image is provided as an output by the acquisition sequence, it is used as an input to offline calculation of a T1map using a dictionary lookup approach. However, complex data is needed for an accurate calculation of the UNIT1 image, which is not commonly provided by the stock sequence. Instead, the magnitude and phase images are exported. Please see the relevant discussion at qMRLab issue #255.

Therefore, the UNIT1 image provided by the scanner SHOULD be stored under the anat in a raw BIDS dataset along with the MP2RAGE file collection and to be used as the primary input for quantifying a T1map.

If an additional UNIT1 image is calculated offline, then the output MUST be stored in a derivative BIDS dataset with necessary provenance information.

NumberShots metadata field

Note that the type of NumberShots field can be either a number or an array of numbers.

  • If a single number is provided, this should correspond to the number of SlicesPerSlab or ReconMatrixPE. However, in this case, SlicePartialFourier or PartialFourierPE fraction is needed to calculate the number of partitions before and after of the k-space center to calculate a T1 map.

  • If before/after calculation is performed during the BIDS conversion of the MP2RAGE data, then the value of NumberShots metadata field can be given as a 1X2 array, with first entry corresponding to before and the second to the after.

Formula:

If NumberShots is an array of numbers such that "NumberShots": [before, after], the values of before and after are calculated as follows:

before = SlicesPerSlab*(SlicePartialFourier - 0.5)
after  = SlicesPerSlab/2

See this reference implementation.

Other metadata fields

The value of the RepetitionTimeExcitation field is not commonly found in the DICOM files. When accessible, the value of EchoSpacing corresponds to this metadata. When not accessible, 2 X EchoTime can be used as a surrogate.

Further information about other MP2RAGE qMRI protocol fields can be found in the qMRLab documentation.