Fetching atlases and annotations

This example demonstrates how to use neuromaps.datasets to fetch atlases and annotations.

Much of the functionality of the neuromaps toolbox relies on the atlases and atlas files provided with it. In many cases these atlases are fetched “behind-the-scenes” when you call functions that depend on them, but they can be accessed directly.

There is a general purpose neuromaps.datasets.fetch_atlas() function that can fetch any of the atlases provided with neuromaps:

from neuromaps import datasets

fslr = datasets.fetch_atlas(atlas='fslr', density='32k')
print(fslr.keys())

Downloading data from https://files.osf.io/v1/resources/4mw3a/providers/osfstorage/60b684b53a6df1021bd4df2d ...
 ...done. (3 seconds, 0 min)
Extracting data from /home/runner/neuromaps-data/599046a594e0e45c04e90291c2348cbe/fsLR32k.tar.gz..... done.
dict_keys(['midthickness', 'inflated', 'veryinflated', 'sphere', 'medial', 'sulc', 'vaavg'])

The values corresponding to the keys of the atlas dictionary are length-2 lists containing filepaths to the downloaded data. All surface atlas files are provide in gifti format (whereas MNI files are in gzipped nifti format).

You can load them directly with nibabel to confirm their validity:

import nibabel as nib
lsphere, rsphere = fslr['sphere']
lvert, ltri = nib.load(lsphere).agg_data()
print(lvert.shape, ltri.shape)

(32492, 3) (64980, 3)

The other datasets that are provided with neuromaps are annotations (i.e., brain maps!). While we are slowly making more and more of these openly available, for now only a subset are accessible to the general public; these are returned by default via datasets.available_annotations().

annotations = datasets.available_annotations()
print(f'Available annotations: {len(annotations)}')

Available annotations: 86

The available_annotations() function accepts a number of keyword arguments that you can use to query specific datasets. For example, providing the format=’volume’ argument will return only those annotations that are, by default, a volumetric image:

volume_annotations = datasets.available_annotations(format='volume')
print(f'Available volumetric annotations: {len(volume_annotations)}')

Available volumetric annotations: 47

There are a number of keyword arguments we can specify to reduce the scope of the annotations returned. Here, source specifies where the annotation came from (i.e., a dataset from a manuscript or a data repository or toolbox), desc refers to a brief description of the annotation, space clarifies which space the annotation is in, and den (specific to surface annotations) clarifies the density of the surface on which the annotation is defined:

annot = datasets.available_annotations(source='abagen', desc='genepc1',
                                       space='fsaverage', den='10k')
print(annot)

[('abagen', 'genepc1', 'fsaverage', '10k')]

Annotations also have tags to help sort them into categories. You can see what tags can be used to query annotations with the available_tags() functions:

tags = datasets.available_tags()
print(tags)

['ASL', 'MEG', 'MRI', 'PET', 'fMRI', 'functional', 'genetics', 'meta-analysis', 'metabolism', 'receptors', 'resteyesopen', 'structural']

Tags can be used as a keyword argument with available_annotations(). You can supply either a single tag or a list of tags. Note that supplying a list will only return those annotations that match ALL supplied tags:

fmri_annotations = datasets.available_annotations(tags='fMRI')
print(fmri_annotations)

[('margulies2016', 'fcgradient01', 'fsLR', '32k'), ('margulies2016', 'fcgradient02', 'fsLR', '32k'), ('margulies2016', 'fcgradient03', 'fsLR', '32k'), ('margulies2016', 'fcgradient04', 'fsLR', '32k'), ('margulies2016', 'fcgradient05', 'fsLR', '32k'), ('margulies2016', 'fcgradient06', 'fsLR', '32k'), ('margulies2016', 'fcgradient07', 'fsLR', '32k'), ('margulies2016', 'fcgradient08', 'fsLR', '32k'), ('margulies2016', 'fcgradient09', 'fsLR', '32k'), ('margulies2016', 'fcgradient10', 'fsLR', '32k'), ('mueller2013', 'intersubjvar', 'fsLR', '164k'), ('neurosynth', 'cogpc1', 'MNI152', '2mm')]

Once we have an annotation that we want we can use the neuromaps.datasets.fetch_annotation() to actually download the files. This has a very similar signature to the available_annotations() function, accepting almost all the same keyword arguments to specify which annotations are desired.

Here, we’ll grab the first principal component of gene expression across the brain (from the Allen Human Brain Atlas):

abagen = datasets.fetch_annotation(source='abagen', desc='genepc1')
print(abagen)

Downloading data from https://files.osf.io/v1/resources/4mw3a/providers/osfstorage/60c2290118f70b01fca797eb ...
 ...done. (1 seconds, 0 min)
Downloading data from https://files.osf.io/v1/resources/4mw3a/providers/osfstorage/60c228f5f3ce9401fa24e521 ...
 ...done. (1 seconds, 0 min)

[References] Please cite the following papers if you are using this data:

  For {'source': 'abagen', 'desc': 'genepc1', 'space': 'fsaverage', 'den': '10k'}:
  [primary]:
    Michael J Hawrylycz, Ed S Lein, Angela L Guillozet-Bongaarts, Elaine H Shen, Lydia Ng, Jeremy A Miller, Louie N Van De Lagemaat, Kimberly A Smith, Amanda Ebbert, Zackery L Riley, and others. An anatomically comprehensive atlas of the adult human brain transcriptome. Nature, 489(7416):391, 2012.
    Ross D Markello, Aurina Arnatkeviciute, Jean-Baptiste Poline, Ben D Fulcher, Alex Fornito, and Bratislav Misic. Standardizing workflows in imaging transcriptomics with the abagen toolbox. eLife, 10:e72129, 2021.
  [secondary]:

['/home/runner/neuromaps-data/annotations/abagen/genepc1/fsaverage/source-abagen_desc-genepc1_space-fsaverage_den-10k_hemi-L_feature.func.gii', '/home/runner/neuromaps-data/annotations/abagen/genepc1/fsaverage/source-abagen_desc-genepc1_space-fsaverage_den-10k_hemi-R_feature.func.gii']

Notice that the returned annotation abagen is a dictionary. We can subset the dictionary with the appropriate key or, if we know that our query is going to return only one annotation, also provide the return_single=True argument to the fetch call:

abagen = datasets.fetch_annotation(source='abagen', desc='genepc1',
                                   return_single=True)
print(abagen)

[References] Please cite the following papers if you are using this data:

  For {'source': 'abagen', 'desc': 'genepc1', 'space': 'fsaverage', 'den': '10k'}:
  [primary]:
    Michael J Hawrylycz, Ed S Lein, Angela L Guillozet-Bongaarts, Elaine H Shen, Lydia Ng, Jeremy A Miller, Louie N Van De Lagemaat, Kimberly A Smith, Amanda Ebbert, Zackery L Riley, and others. An anatomically comprehensive atlas of the adult human brain transcriptome. Nature, 489(7416):391, 2012.
    Ross D Markello, Aurina Arnatkeviciute, Jean-Baptiste Poline, Ben D Fulcher, Alex Fornito, and Bratislav Misic. Standardizing workflows in imaging transcriptomics with the abagen toolbox. eLife, 10:e72129, 2021.
  [secondary]:

['/home/runner/neuromaps-data/annotations/abagen/genepc1/fsaverage/source-abagen_desc-genepc1_space-fsaverage_den-10k_hemi-L_feature.func.gii', '/home/runner/neuromaps-data/annotations/abagen/genepc1/fsaverage/source-abagen_desc-genepc1_space-fsaverage_den-10k_hemi-R_feature.func.gii']

And that’s it! This example provided a quick overview on how to fetch the various atlases and datasets provided with neuromaps. For more information please refer to the API reference.