diff --git a/README.md b/README.md index 6474f470a1..40c29b6b81 100644 --- a/README.md +++ b/README.md @@ -30,6 +30,7 @@ BIDS currently supports the following data modalities with more to come in the f - microscopy - NIRS - motion +- MRS # Formatting your data with BIDS diff --git a/mkdocs.yml b/mkdocs.yml index 114e05c135..cfbdba2a49 100644 --- a/mkdocs.yml +++ b/mkdocs.yml @@ -19,6 +19,7 @@ nav: - Microscopy: modality-specific-files/microscopy.md - Near-Infrared Spectroscopy: modality-specific-files/near-infrared-spectroscopy.md - Motion: modality-specific-files/motion.md + - Magnetic Resonance Spectroscopy: modality-specific-files/magnetic-resonance-spectroscopy.md - Derivatives: - BIDS Derivatives: derivatives/introduction.md - Common data types and metadata: derivatives/common-data-types.md diff --git a/src/appendices/cross-modality-correspondence.md b/src/appendices/cross-modality-correspondence.md index c71a1c73f5..6dba22b325 100644 --- a/src/appendices/cross-modality-correspondence.md +++ b/src/appendices/cross-modality-correspondence.md @@ -12,3 +12,10 @@ The reason for this is that the MRI needs to be corrected for nonlinear gradient in order to fit the accompanying PET scans for co-registration (Knudsen et al. 2020, [doi:10.1177/0271678X20905433](https://doi.org/10.1177/0271678X20905433); Norgaard et al. 2019, [doi:10.1016/j.neuroimage.2019.05.055](https://doi.org/10.1016/j.neuroimage.2019.05.055)). + +## MRS-MRI correspondence + +It is typical to acquire high-resolution 3D anatomical MR images alongside MRS data for +voxel/slab placement, co-registration, and partial-volume tissue correction of metabolite concentrations. +To avoid incorrectly matching an MRS dataset with a corresponding anatomical MR image, +it is RECOMMENDED that the field `AnatomicalImage` be included in the MRS sidecar JSON files. diff --git a/src/introduction.md b/src/introduction.md index 4315720e3a..a07b654b2c 100644 --- a/src/introduction.md +++ b/src/introduction.md @@ -184,6 +184,10 @@ For example: PsyArXiv. [doi:10.31234/osf.io/w6z79](https://doi.org/10.31234/osf.io/w6z79) +#### MRS + +- (publication forthcoming) + ### Research Resource Identifier (RRID) BIDS has also a diff --git a/src/modality-specific-files/magnetic-resonance-spectroscopy.md b/src/modality-specific-files/magnetic-resonance-spectroscopy.md new file mode 100644 index 0000000000..22ec6540e0 --- /dev/null +++ b/src/modality-specific-files/magnetic-resonance-spectroscopy.md @@ -0,0 +1,260 @@ +# Magnetic Resonance Spectroscopy + +Support for Magnetic Resonance Spectroscopy (MRS) was developed as a +[BIDS Extension Proposal](../extensions.md#bids-extension-proposals). +Please see [Citing BIDS](../introduction.md#citing-bids) +on how to appropriately credit this extension when referring to it in the +context of the academic literature. + +!!! example "Example datasets" + + Several [example MRS datasets](https://github.com/bids-standard/bids-examples/pull/425) have + been formatted using this specification and can be used for practical guidance when curating a new + dataset. + +## MRS data + + +{{ MACROS___make_filename_template("raw", datatypes=["mrs"]) }} + +MRS is a spectroscopic technique based on the phenomenon of nuclear magnetic resonance +that allows for the noninvasive detection and quantification of molecules in biochemical samples, such as brain tissue. +It can be conducted in humans using conventional MRI systems. + +Due to the diversity in manufacturers' MRS data file formats, source data MUST be converted into the +[NIfTI-MRS format](https://wtclarke.github.io/mrs_nifti_standard/) (`*.nii[.gz]`) ([doi:10.1002/mrm.29418](https://doi.org/10.1002/mrm.29418)). +This format is based on the NIfTI framework and is designed to accommodate the nuances of raw MRS data. +All necessary information to parse this `*.nii[.gz]` file (for example, spectrometer frequency, echo time, +repetition time, and so on) are stored in a JSON header extension. +Conversion of proprietary MRS file formats to NIfTI-MRS and extraction of some (but not all) BIDS-compliant metadata can be performed +using [spec2nii](https://github.com/wtclarke/spec2nii). +Note that the "rawness" of data stored in the NIfTI-MRS file will depend on the format of the source data. +It is RECOMMENDED that users export their source data from the scanner in an appropriately raw format prior to conversion. + +For MRSI data, "raw" signifies spatially reconstructed data (that is, data in image space rather than (*k*,*t*)-space), +given the complexity and diversity of sampling approaches. +Note that NIfTI-MRS is not designed to store data that has not been spatially reconstructed. + +Regarding source data, each manufacturer has its own file format (sometimes multiple formats) for exporting MRS data from +the MRI scanner console for offline processing. +GE exports a P-file (`*.7`) that stores unprocessed, un-coil-combined data with metadata embedded +in a proprietary data header. +Philips has multiple export formats, the most common being the SDAT/SPAR format. +The `*.sdat` file contains either each coil-combined transient stored separately +or all transients summed into a signal average. +The `*.spar` file is a plaintext file describing acquisition parameters. +It is also possible to export raw data as `*.data`/`*.list` or DICOM files. +Siemens scanners allow data export in four formats: i) a proprietary DICOM-structured file known as IMA (`*.ima`); +ii) a conventional DICOM MR Spectroscopy Storage format (`*.dcm`); iii) RDA (`*.rda`), +a proprietary file format with a text-formatted header followed by the binary data points; +and iv) TWIX (`*.dat`), a proprietary file format designed for storing unreconstructed, unprocessed MRS data +from each individual coil element. +The IMA, DICOM MRS, and RDA formats are typically used to export reconstructed and processed data; +however, the sequence designer may choose to also allow the export of un-averaged transients +or data from individual coil elements. +Bruker data are are exported as two binary files: one file stores each transient separately, +while the other stores the sum of the transients. +A separate plaintext file stores the sequence name, voxel position, voxel orientation, and other metadata. +All of these files are considered source data and, if present, MUST be stored in the +[`sourcedata`](../common-principles.md#source-vs-raw-vs-derived-data) directory. + +### Single-voxel spectroscopy and MRS imaging + +| **Name** | **`suffix`** | **Description** | +| ---------------------------------------- | ------------ | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | +| Single-voxel spectroscopy | svs | MRS acquisitions where the detected MR signal is spatially localized to a single volume. | +| Magnetic resonance spectroscopic imaging | mrsi | MRS acquisitions where additional imaging gradients are used to detect the MR signal from 1, 2, or 3 spatial dimensions. | +| Unlocalized spectroscopy | unloc | MRS acquisitions run without localization. This includes signals detected using coil sensitivity only. | +| Concentration or calibration reference | mrsref | An MRS acquisition collected to serve as a concentration reference for absolute quantification or as a calibration reference for preprocessing (for example, eddy-current correction). | + +A major distinction between MRS acquisitions is whether the acquisition technique probes spectral +information from a single volume (single-voxel spectroscopy, SVS) or encodes this information along +1, 2, or 3 spatial dimensions resulting in multiple sub-volumes (MRS imaging, MRSI). +To avoid confusion, the suffixes `svs` and `mrsi` MUST be used to distinguish the two techniques. +For cases where localization is not used, the suffix `unloc` MUST be used. + +Furthermore, it is common to acquire an additional MRS dataset that may serve as a reference for +scaling metabolite signal levels (for example, to obtain concentrations) and/or for preprocessing steps (such as +eddy-current correction, RF coil combination, phasing, and frequency calibration). +This could be either an external reference (for example, a phantom or a synthetic signal) or, more typically, +an internal tissue water reference. +For such datasets, the suffix `mrsref` MUST be used. +Should multiple references exist for a given dataset, the user MAY use the `acq-