Ultrahigh resolution 3D cytoarchitectonic map of the human Corpus geniculatum mediale (CGM, MGB, MGBv, MGBd, MGBm, metathalamus)

This dataset contains cytoarchitectonic maps of the ventral (MGBv), dorsal (MGBd) and medial (MGBm) subdivisions of the human Corpus geniculatum mediale (CGM, MGB, metathalamus, from here on called MGB) in the BigBrain dataset, which has a spatial resolution of 20 micrometres isotropic (Amunts et al., 2013). Mappings were created based on delineations of the MGB subdivisions on 57 sections of the BigBrain. Deep Convolutional Neural networks were subsequently trained on these delineations to segment the subdivisions on sections in-between. Resulting maps are available on every section and were transformed to the 3D reconstructed BigBrain space. Individual sections were used to assemble a 3D volume of the area. Low quality results were replaced by manual expert delineations in single cases. The volume was then smoothed using median filter and largest connected components were identified to remove false positive results. The dataset consists of a HDF5 file containing the volume in RAS dimension ordering (20-micron isotropic resolution, dataset “volume”) and an affine transformation matrix (dataset “affine”). An additional dataset “interpolation_info” contains an integer vector for each section which indicates if a section was replaced by interpolation.