Standardizing human brain parcellations

Abstract

Using brain atlases to localize regions of interest is a requirement for making neuroscientifically valid statistical inferences. These atlases, represented in volumetric or surface coordinate spaces, can describe brain topology from a variety of perspectives. Although many human brain atlases have circulated the field over the past fifty years, limited effort has been devoted to their standardization. Standardization can facilitate consistency and transparency with respect to orientation, resolution, labeling scheme, file storage format, and coordinate space designation. Our group has worked to consolidate an extensive selection of popular human brain atlases into a single, curated, open-source library, where they are stored following a standardized protocol with accompanying metadata, which can serve as the basis for future atlases. The repository containing the atlases, the specification, as well as relevant transformation functions is available in the neuroparc OSF registered repository or https://github.com/neurodata/neuroparc .

Fig. 1

A comparison of the regions present in the major atlases available in Neuroparc. These visualizations were made using MIPAV tri-planar views on the same slice numbers. Each atlas shows a cross-section in each of the canonical orthogonal planes (H = Horizontal, S = Sagittal, C = Coronal). For most atlases, the slice numbers were (90, 108, 90). There are a few exceptions for visualization purposes: JHU: (90, 108, 109), Slab907: (95, 104, 95), Slab1068: (93, 105, 93)^,,,,,,,,–.

Fig. 2

Dice Score Map between the Yeo-17 Networks atlas and the 300 parcellation Schaefer atlas. In a Dice Score Map, the larger the Dice score, the larger the percentage of overlap. Due to Schaefer’s larger quantity of ROIs, several different ROIs overlap a single Yeo-17 ROI. In this Dice Map, the 0 ROI for Yeo-17 represents the background of the image, or the empty space in the image. This ROI not having a Dice value of 0 indicates that both atlases don’t cover the same amount of brain volume.

Fig. 3

The adjusted mutual information between atlases contained within Neuroparc. Atlases that were generated from the same algorithm using different parameters, such as Yeo, Slab, Schaefer, and DS have an expected high amount of mutual information.

Fig. 4

An example JSON file rubric for storing atlas metadata. Metadata stored in brackets (“color”, “description”, etc.) is optional but encouraged.