niatlas

Atlas classes and methods for neuroimaging

This is a placeholder repository for potential development of nibabel-like python package that is intended to make the management of Atlases for neuroimaging seamless, in all their diverse use-cases. We will collect and link to various relevant resources to facilitate this discussion and project development.

As I currently envision it, a decent niatlas package would offer the following features (consider it a wishlist):

• easy access (including I/O) to all the popular atlases just by their name, both volumetric- and surface-based.

• a well-defined data structure that provides, not only the parcellations, but also all the relevant meta-data, such as

  • the source of atlas, in terms of modalities and tha processes that generated it
  • methods defining the parcellation,
  • number, names and centroids of ROIs, along with resolution and dimensions
  • whether it is intended to be used as a volumetric or surface atlas,
  • domain tags that identify which age-group this atlas would be ideal for, along with other info related to target population
  • etc

• several convenience methods to perform the common operations on atlases including but not limited to

  • computing ROI-based statistics
  • masking operations

• Methods to obtain different variations of the same atlas e.g.

  • resampling the parcellation to a different resolution, or to different dimensions (that respects the internal parcellations)
  • scale control in terms of number or size of ROIs i.e. methods for subdividing or clustering existing ROIs
  • conversion to different spaces, such as between volumetric and surface-oriented spaces
  • conversion between atlas- and subject-spaces

• visualization routines for all the common analyses needs
• easy integration and high interoperability with popular tools and ecosystems

Some prior discussion on potential data structures for Atlas object and uniform access to parcellations at nilearn here

Prior Art (software)

• nilearn fetch_{atlas} utilities
• nilabels : tools to automate simple manipulations and measurements of medical image segmentations
• AtlasReader to generate coordinate tables and region labels from statistical MRI images : https://github.com/miykael/atlasreader
• pysurfer visualization tool: https://pysurfer.github.io/auto_examples/index.html
• python package for subparcellation of fsaverage etc: https://github.com/miykael/parcellation_fragmenter
• shell scripts to move from atlas space to subject space https://github.com/faskowit/multiAtlasTT

Resources - atlas collections

• Different atlases in MNI space: http://www.lead-dbs.org/helpsupport/knowledge-base/atlasesresources/cortical-atlas-parcellations-mni-space/

• Different atlases from Thomas Yeo's lab

  • set 1: https://github.com/ThomasYeoLab/CBIG/tree/master/stable_projects/brain_parcellation
  • set 2: https://github.com/ThomasYeoLab/CBIG/tree/master/data

• BALSA (Brain Analysis Library of Spatial maps and Atlases) is a database for hosting and sharing neuroimaging and neuroanatomical datasets for human and primate species.
• Brain signature patterns, atlases of regions, and meta-analysis masks from Tor Wager's lab: https://github.com/canlab/Neuroimaging_Pattern_Masks