Marmoset Brain Connectivity Atlas

Supplementary material

The web page http://www.marmosetbrain.org/whole_brain_cb_maps contains the most recent versions of the supplementary materials for the article:

Distribution of calbindin-positive neurons
across areas and layers of the marmoset cerebral cortex

Abstract

The diversity of the mammalian cerebral cortex demands technical approaches to map the spatial distribution of neurons with different chemical identities. This issue is magnified in the case of the primate cortex, characterized by a large number of areas with distinctive cytoarchitectures. To date, no full map of the distribution of cells expressing a specific protein has been reported for the cortex of any primate. Here we have charted the 3-dimensional distribution of neurons expressing the calcium-binding protein calbindin (CB⁺ neurons) across the entire marmoset cortex, using a combination of immunohistochemistry, automated cell identification, computerized reconstruction, and cytoarchitecture-aware registration. CB⁺ neurons formed a heterogeneous population, which together corresponded to 10-20% of the cortical neurons. They occurred in higher proportions in areas corresponding to low hierarchical levels of processing, such as sensory cortices. Although CB⁺ neurons were concentrated in the supragranular and granular layers, there were clear global trends in their laminar distribution. For example, their relative density in infragranular layers increased with hierarchical level along sensorimotor processing streams, and their density in layer 4 was lower in areas involved in sensorimotor integration, action planning and motor control. These results reveal new quantitative aspects of the cytoarchitectural organization of the primate cortex and demonstrate an approach to mapping the full distribution of neurochemically distinct cells throughout the brain, which is readily applicable to most other mammalian species.

Supporting Table S1: Source dataset for reproducing analyses presented in Figure 3-7.
Supporting Table S2: Source dataset for analyses of the variance in densities.
Supporting Table S3: Source dataset for valuation of the registration accuracy.
Supporting File S1: Reference card with the full names, abbreviations, color-codes and flatmap locations of the individual cortical areas, groups of areas, and the cytoarchitectural categories of lamination.

For each of the three cases (CJ1741, CJ200L, and CJ205) the following datasets are available:

• {case_id}_CB_density_map.nii.gz
  Three-dimensional image of the density map of CB⁺ neurons (in mm^-3).
• {case_id}_3d_reconstruction.nii.gz
  Three-dimensional image of the calbindin sections. The image is compatible with the density map.
• {case_id}_segmentation_into_areas.nii.gz
  Segmentation of the cortex into areas areas. To be used together with the atlas_labels.txt label description file.
• {case_id}_segmentation_into_layers.nii.gz
  Segmentation of the cortex into supragranular (I-III), granular (IV), and infragranular (VI) cell layers.
• {case_id}_mask.nii.gz
  Mask used to calculate the values reported in Supplemental Table 1. The mask takes into account the artifacts and other parts of the tissue (i.e. Layer 1) that were decided to be excluded from the analyses.
• atlas_labels.txt
  A file containing label descriptions for the segmentation into cortical areas.

The dataset used for training and validation of the Unet CNN is available here. It contains:

• The training dataset contains 196 image strips (3,429 counting boxes) used to train the Unet CNN model.
• The validation dataset contains 54 image strips (603 counting boxes) used to evaluate the model’s performance.
• The benchmark dataset contains 81 image strips (939 counting boxes) used to compare the Unet CNN against multiple human raters.
• Each dataset contains a series of TIFF files containing the image strips extracted from the high-resolution microscopic (0.4974 µm per pixel) images. They are accompanied by corresponding SVG files containing annotations (definitions of counting boxes and individual cells).

The Python scripts for training the Unet CNN and the input NumPy datasets are available for download here.

How to view the NIFTI files?:

The most convenient way of viewing the atlas is to use the ITKSnap software (http://www.itksnap.org; Yushkevich, et al. (2006)).

You can watch this video:
https://youtu.be/AuFavplRvsA?t=2110

Or use one of the following links to get some more information on using the ITKSnap softare:

• https://www.youtube.com/watch?v=BvQyjk1etLU&list=PL6dt1cs7som7I4HSi1rPquJOjrUlnLqtD
• http://www.itksnap.org/pmwiki/pmwiki.php?n=Videos.SNAP3
• http://www.itksnap.org/pmwiki/pmwiki.php?n=Documentation.SNAP3

Licensing

The materials available on this website are licensed under the Creative Commons Attribution 4.0 International (CC BY) License.