diff --git a/src/atlas.md b/src/atlas.md index 6a2297b97f..10a68845d5 100644 --- a/src/atlas.md +++ b/src/atlas.md @@ -1,183 +1,57 @@ -# BIDS-Atlas +# Explanation In the following we describe how an [atlas](schema/objects/entities.yaml#atlas) can be shared within BIDS. -We describe a broad set of atlases and use cases thereof. +Broadly we define an atlas as a reference quantity sampled to a template. +With this definition, an atlas can be a parcellation, a segmentation, or a quantitative map. -More specifically, this entails providing and referring to existing atlas datasets, -describing atlases that were newly derived within an analysis, -and providing information for derivatives that were obtained through them. -The first would comprise (publicly) available atlases, for example, +We aim to cover three main use cases with the given definition: +1. Generating/sharing atlas(es) to be published and used by others. +2. Using a shared atlas and transforming/resampling it into new template spaces or subject spaces. +3. Creating subject-specific atlases derived from the corresponding subject’s data. + +These three use cases cover the majority of scenarios where atlases are used in neuroimaging research. + +The first use case would comprise (publicly) available atlases, for example, Destrieux et al. ([doi.org/10.1016/j.neuroimage.2010.06.010](https://doi.org/10.1016/j.neuroimage.2010.06.010)), AAL ([doi.org/10.1006/nimg.2001.0978](https://doi.org/10.1006/nimg.2001.0978)), Yeo ([doi.org/10.1152/jn.00338.2011](https://doi.org/10.1152/jn.00338.2011)) and JHU DTI-based white-matter atlases ([eBook ISBN: 9780080456164](https://shop.elsevier.com/books/mri-atlas-of-human-white-matter/mori/978-0-444-51741-8) -and [doi.org/10.1016/j.neuroimage.2007.07.053](https://doi.org/10.1016/j.neuroimage.2007.07.053)), -while the second would include atlases obtained through analyses within a dataset at hand, +and [doi.org/10.1016/j.neuroimage.2007.07.053](https://doi.org/10.1016/j.neuroimage.2007.07.053)). +The second use case would comprise transforming/resampling the publicly available atlases into new template spaces or subject spaces for further analysis. +For example, resampling the AAL atlas from MNI152NLin6Asym space to a subject’s native T1w space. +The third use case would include atlases obtained through analyses within a dataset at hand, for example, resting-state networks and functional localizers. -Importantly, the latter can also be utilized as existing atlases if made available. -The third would entail referencing an atlas and its properties used to derive, -for example, a parcellated time series or a connectivity matrix. - -## Atlas as new DatasetType -Here we introduce an additional value to the `DatasetType field` of `dataset_description.json`. -If a dataset declares its DatasetType to be [atlas](schema/objects/entities.yaml#atlas), -the top-level directories MUST be `atlas-` instead of `sub-`. -This will allow sharing existing atlases as stand-alone datasets, -validating them via the [BIDS validator](https://github.com/bids-standard/bids-validator) and enabling their integration as sub-datasets of other BIDS datasets. - -## File formats for the raw data -BIDS-Atlas aims to describe brain atlases via three REQUIRED files. -They entail the atlas itself (for example, in .nii, .nii.gz, .gii, or .tsv), -a file indexing/labeling each node in the atlas (in .tsv) and a file containing exhaustive meta-data about the atlas (in .json). - -The usage of [_desc-](schema/objects/entities.yaml#desc) is generally discouraged but should be evaluated on a case by case basis in order to keep this identifier available for necessary cases. -Specifically, this refers to the atlas at hand and potential different versions thereof. -As a rule of thumb, BIDS-Atlas proposed to evaluate and consider how many versions across how many levels of versions an atlas is (commonly) provided and used in. -If there is only one version, as in "release", of an atlas -and no sub-versions, as in "different parcel numbers" (or comparable), -[_desc-](schema/objects/entities.yaml#desc) is most likely not expected and/or required. -An example for [_desc-](schema/objects/entities.yaml#desc) in such a use case would be -to indicate a subset of parcels and would entail the addition of -the [_mask-](schema/objects/entities.yaml#mask) extension, -for example, providing/using the "postcentral gyrus" region of the [Destrieux atlas](doi:10.1093/cercor/bhg087.) would result in the following file name: -`atlas-Destrieux_space-MNI152NLin6Asym_res-2_desc-PostCentralGyrus_mask.nii.gz`. -Similar to the above case, if there are multiple versions, as in "release", -of an atlas and no sub-versions, as in "different parcel numbers" (or comparable), -[_desc-](schema/objects/entities.yaml#desc) is also most likely not expected and/or required, -as the version(s) (release(s)) should be indicated via `atlas-`. -For example, the different versions of the [AAL parcellation](http://www.gin.cnrs.fr/AAL-217?lang=en) -would result in the following file names: -`atlas-AAL1_space-MNI152NLin6Asym_res-2.nii.gz`, `atlas-AAL2_space-MNI152NLin6Asym_res-2.nii.gz` and `atlas-AAL3_space-MNI152NLin6Asym_res-2.nii.gz`. -Given an atlas has only one version, as in "release", and multiple sub-versions -as in "different parcel numbers" (or comparable) -[_desc-](schema/objects/entities.yaml#desc) is considered appropriate. -For example, when indicating information pertaining to the probabilistic nature of an atlas -such as the [_probseg](schema/objects/entities.yaml#probseg) version of -the [Harvard-Oxford parcellation](https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/Atlases): -`atlas-HarvardOxford_res-2_desc-th25_probseg.nii.gz`. -In cases where an atlas has multiple versions, as in "release", and sub-versions, -as in "different parcel numbers" (or comparable), -the version(s) (release(s)) should be indicated via `atlas-` as outlined above and the sub-versions should be indicated via [_desc-](schema/objects/entities.yaml#desc). -For example, different versions and sub-versions of the [Schaefer parcellation](https://github.com/ThomasYeoLab/CBIG/blob/master/stable_projects/brain_parcellation/Schaefer2018_LocalGlobal/Parcellations/Updates/Update_20190916_README.md) would be denoted as follows: -`atlas-Schaefer2018_space-MNI152NLin6Asym_res-2_desc-400Parcels7Networks.nii.gz`, -`atlas-Schaefer2018_space-MNI152NLin6Asym_res-2_desc-400Parcels17Networks.nii.gz`, -`atlas-Schaefer2022_space-MNI152NLin6Asym_res-2_desc-800Parcels7Networks.nii.gz` -and `atlas-Schaefer2022_space-MNI152NLin6Asym_res-2_desc-800Parcels17Networks.nii.gz`. - -Importantly, already existing BIDS entities should be used to indicate certain aspects of an atlas, -instead of [_desc-](schema/objects/entities.yaml#desc), for example, -[hemi](schema/objects/entities.yaml#hemi) to denote a given hemisphere -and [_dseg](schema/objects/entities.yaml#dseg)/[_probseg](schema/objects/entities.yaml#probseg) -to denote deterministic and probabilistic atlases respectively. - -However, as mentioned above, these are general guidelines and the exact implementation should be evaluated on a case by case basis, with deviations following common BIDS principles being permitted. - -## Directory Structure -BIDS-Atlas focuses on the utilization of atlases while also allowing their sharing. -Thus, atlases are either stored within a dedicated `atlas` directory at the BIDS root directory -(comparable to the `code` directory), -such files are the non-altered/original atlases or within a given directory under `derivatives`. -In the second case, atlases are altered, -derived or applied and thus multiple use cases have to be distinguished as indicated further below. - -### Representing an atlas as a dataset - -The first option refers to atlases that were not altered, for example, -via spatial transformations and/or resampling or applied to data -and thus their initial inclusion/utilization in a given dataset. -If there is only this form of atlases (that is, the tool used), -they are always shared at the root directory and everything else is under derivatives. -This allows validating any - -```Text -/atlas/ -``` - -using the [BIDS validator](https://github.com/bids-standard/bids-validator). -Importantly, only this case follows the same directory structure as -[sub](schema/objects/entities.yaml#sub), that is, one dedicated directory for each atlas within a given dataset. -The default way of storage of the non-altered atlas at the root directory looks like this: - -```Text -/atlas/atlas-